International Journal of Computer Applications in Technology (105 papers in press)
Simulation and visualisation approach for accidents in chemical plants
by Feng Ting-Fan, Tan Jing, Liu Jin, Deng Wensheng
Abstract: A new general approach to lay the foundation for building a more effective and real-time evacuation system for accidents in chemical plants is presented. In this work, we build the mathematical models and realise automatic grid generating based on the physical models stored in advance with several algorithms in jMonkeyEngine environment. Meanwhile, the results of the simulation data through finite difference method (FDM) are visualised coupling with the physical models. Taking fire as an example, including fire with single and multiple ignition sources, shows the feasibility of the presented approach. Furthermore, a coarse alarm and evacuation system from fire have been developed with a multiple SceneNode and roam system, which also includes the making and importing of the physical models. However, to improve the accuracy of the mathematical models, adaptability and refinement of the grids and universality of the evacuation system is the direction of efforts.
Keywords: simulation; chemical accidents; alarm and evacuation system; jMonkeyEngine.
Detecting occluded faces in unconstrained crowd digital pictures
by Chandana Withana, S. Janahiram, Abeer Alsadoon, A.M.S. Rahma
Abstract: Face detection and recognition mechanisms, a concept known as face detection, are widely used in various multimedia and security devices. There are significant numbers of studies into face recognition, particularly for image processing and computer vision. However, there remain significant challenges in existing systems owing to limitations behind algorithms. Viola Jones and Cascade Classifier are considered the best algorithms from existing systems. They can detect faces in an unconstrained crowd scene with half and full face detection methods. However, limitations of these systems are affecting accuracy and processing time. This project proposes a solution called Viola Jones and Cascade (VJaC), based on the study of current systems, features and limitations. This system considered three main factors: processing time, accuracy and training. These factors are tested on different sample images, and compared with current systems.
Keywords: face detection; unconstrained crowd digital pictures; face recognition.
Hierarchical smart Routing protocol for wireless sensor networks
by Amine Kardi, Rachid Zagrouba
Abstract: This paper proposes a new hierarchical routing protocol dedicated to wireless sensor networks. Several solutions have been proposed, but an optimal one, which can be applied in different situations, still remains non-existent. Our proposal takes advantage of existing solutions and proposed a new radial cluster head (CH) selection algorithm to ensure the QoS, good load balancing and enhance the network lifetime. A performance analysis of our proposal with LEACH, Mod_LEACH and M-GEAR protocols is presented in this paper comparing these protocols using different metrics, such as network lifetime, throughput, stability and instability periods, remaining energy, network overhead. The simulations show that our solution outperforms existing solutions owing to the radial architecture proposed for CH selection and the use of two transmission levels for intra- and inter-cluster communication. Obtained results show that our proposal improves the network lifetime by 113% compared with LEACH protocol and optimises the energy consumption by more than 62% in comparison with M-GEAR.
Keywords: routing protocol; wireless sensor networks; hierarchical routing protocol; wireless sensor network; cluster head.
Real time sign languages character recognition
by Sari Awwad, Sahar Idwan, Hasan Gharaibeh
Abstract: Deaf people face many challenges to communicate with the hearing world. Many studies and industrial solutions have come up with interpretations from sign language to normal text, but most of them are limited either to static images for letters or to a static animated character playing the word in motion. Therefore, this research worked on enhancing the feasible algorithms that have been used, such as image detection, image processing techniques and image translation methods. The contribution consists of two parts, the first one is using our dataset, and the second one is proposing a new solution by extracting surf features after using image filtering based on the existing methods to accelerate the translation process for the long sentences. Experimental results over our dataset show a distinguished accuracy compared with other studies in terms of efficiency time and recognition rate of sign language character recognition.
Keywords: sign language dataset; image filtering; surf features; M-SVM2; sign character recognition.
Variable neighborhood search based algorithm to solve the minimum back-walk-free latency problem
by Ban Bang
Abstract: The Minimum Back-Walk-Free Latency Problem (MBLP) is an extension of the Minimum Latency Problem (MLP). It aims to find a tour with minimum latency while ignoring the back-walking costs. One of the justifications for ignoring the back-walking costs is in applications of message broadcasting for mobile devices. Obviously, MBLP is an NP-hard problem because it is a generation of MLP, hence a metaheuristic needs to be developed to provide near-optimal solutions within a short computation time. However, the main issue of metaheuristics is that they fall into local optima in some cases since the search space of the problem is combinatorial explosion. In order to overcome this drawback, we propose a metaheuristic algorithm that is mainly based on Variable Neighborhood Search (VNS) and shaking techniques to solve the problem. The aim of VNS is to generate diverse neighborhoods by using various neighborhood searches while shaking techniques allow it to guide the search towards an unexplored part of the solution space. Moreover, a technique called Adaptive Memory (AM) is applied to address this problem to balance between the diversity and intensification. The computational results show that the effectiveness and efficiency of the proposed algorithm are comparable with the current state-of-the-art algorithms.
Keywords: MBLP; metaheuristic; VNS.
Pedestrian detection algorithm based on improved single shot MultiBox detector
by Dawei Liu, Shang Gao, Wanda Chi, Di Fan
Abstract: The algorithm proposed in this paper introduces the deeply separable fusion hierarchical feature model into the backbone network of Single Shot MultiBox Detector (SSD) algorithm, which reduces the complexity of the model by using depthwise separable convolution and uses hierarchical structure fusion to enhance features. We add a decode-encoder structure between the backbone network and the additional feature layer to propagate the high-level semantic feature downward through decoding, and then integrate the feature with the shallow local detail feature. At the same time, the encoder is used to transfer more advanced semantic features and output the fused shallow features for pedestrian detection, which improves the classification and detection ability of the model. The miss rate of the improved algorithm in this paper is as low as 9.68% on the INRIA dataset.
Keywords: level feature fusion; decode-encoder; pedestrian detection; semantic features; single shot MultiBox detector; depthwise separable convolution.
An information real-time synchronisation system across different media based on position encoding for mobile computing
by Jingxian Zhou, Guangming Zheng, Baobao Zhang, Chunbo Liu, Zhaojun Gu
Abstract: Information should be spread accurately, synchronised in real time and shared at low cost, especially when distributed and stored in different media concurrently. In the education field, traditional paper writing and multimedia equipment input are often combined to achieve a better teaching result, which is a typical scenario that information must be synchronised across different media. Interactive paper system has been proved to successfully realise information synchronisation in real time. However, existing solutions have some deficiencies, such as poor reproducibility, delay in recognition, or positioning error. What is more serious is that most of the solutions are expensive and not suitable for large-scale promotion. In order to solve the problem of information fusion across different media at low cost, an information real-time synchronisation system based on SONIX Code was designed and implemented innovatively. In this paper, we redesign the size of the code and code unit, and the decode algorithm. Firstly, the design and encoding scheme of dot matrix coded paper based on virtual raster is introduced, and some important parameters including pixel width, feature point size, encoding atom size, and encoding unit size are determined. Then the decoding process is described in detail as five steps: image noise reduction, encoding unit determination, encoding unit rotation correction, encoding unit coordinate decoding and coordinate offset correction. Finally, the system is demonstrated comprehensively, and experiments on an A4 paper were conducted. Besides, a practical test among 350 users has been carried out. All of the results proved this system had good performance on visual effect, writing fluency and positioning accuracy.
Keywords: information parallel storage; position encoding; real-time synchronisation; intelligent hardware system.
Map simulation of a dog's behaviour using population density of probabilistic model
by Jirawat Jiwattanakul, Chawapat Youngjitikornkun, Worapan Kusakunniran, Anuwat Wiratsudakul, Weerapong Thanapongtharm, Kansuda Leelahapongsathon
Abstract: To understand relations in an ecosystem, it is important to gather and characterise information about living creatures and their habitats. This paper focuses on a dog's behaviour, since it is in a great concern where dogs are environmentally and physically correlated to human life. We therefore develop a simulator to demonstrate dogs' behaviors considering their individual and group habits. This could be further used in a prediction process of a widespread occurrence of any possible kind of diseases via dogs' transmission that is created to be purposefully expandable for disease control. The proposed system is developed using Unity and Mapbox SDK, in order to provide a spatial implementation for the simulation of such dogs' movements. In addition, the normal distribution, the kernel density method and the probabilistic model are applied to model movement behaviour, world interaction and behaviour rates, respectively. The simulation is validated on an area of Saibai, located in the north-western of Torres Strait islands, Australia. This reports a median tie-strength of 0.0106 which is slightly different from the value calculated from the GPS information of 0.0113. The relative error is only 6.19%. Then, the simulation is applied to three cities in Thailand, including Khon Kaen, Songkhla, and Prachuap Khiri Khan, with the dog populations of 145.86, 74.70, 76.54, respectively. The calculated tie-strengths are 0.054, 0.039, and 0.042 for the three cities, respectively. They are all higher than the calculated tie-strength of Saibai, owing to the significantly higher average numbers of dogs and group distances. Consequently, the contact rates between any pair of dogs are also higher due to the fact that the dogs' communities in Thailand have larger connections either between dogs or between their communities, when compared with Saibai.
Keywords: dog simulation; Unity; Mapbox SDK; probabilistic model; kernel density method; dog behaviour.
Determining solubility of CO2 in aqueous brine systems via hybrid smart strategies
by Tofigh Sayahi, Afshin Tatar, Alireza Rostami, Amin Anbaz, Khalil Shahbazi
Abstract: Carbon Capture and Storage (CCS) helps the reduction of atmospheric CO2 emission, and increases the oil recovery factor. Solubility of CO2 in water/brine is an important parameter for successful development of the CCS process; however, there exists a large gap to establish a rigorous approach to estimate solubility of CO2 in water/brine. Machine learning as a strong approach can properly tackle such issue in this field of study. In current study, Radial Basis Function Neural Network (RBF-NN) and Least Squares Support Vector Machine (LSSVM) were established for estimation of equilibrium CO2-water/brine solubility as a function of salt molecular weight, temperature, salt molality and pressure. A reliable database was gathered from the open source literature, and split into two groups of testing and training subsets. Optimal structure of the proposed RBF-NN technique and the tuning coefficients of LSSVM model were determined by means of the well-known evolutionary method known as Cuckoo Optimisation Algorithm (COA). Parametric and visual statistics were used to assess the reliability of the suggested models in this study. Accordingly, the proposed approaches here can accurately prognosticate the equilibrium CO2-water/brine solubility with a determination factor (R2) of 0.9966 and average absolute relative deviation percent (AARD%) of 2.5885% for COA-LSSVM, and AARD% = 3.8832% and R2 = 0.9962 for COA-RBF-NN; therefore, the proposed COA-LSSVM gives more accurate results for equilibrium CO2-water/brine solubility. It is understood that the salt molality is most affecting variable on the output prediction based on sensitivity analysis. Williams outliers detection technique reveals that a small number of data (less than 3%) are outliers and also the predictions of the best model proposed here accommodate inside the feasibility region. The proposed methods are of enormous help for engineers and scientists working on CO2 injection into underground formations from both economic and environmental points of view.
Keywords: equilibrium CO2-water/brine; solubility; least squares support vector machine; carbon capture and storage; outliers analysis; radial basis function neural network.
Self-attention-based sentiment analysis with effective embedding techniques
by Soubraylu Sivakumar, Ratnavel Rajalakshmi
Abstract: Corporate companies and organisations use sentiment analysis to find opinions from the review to enhance customer relationship management and to increase the revenue from their product or brand. Sentiment analysis is the process of computationally extracting, identifying and categorising the user-generated opinionated data to determine the writer's attitude towards a particular product, topic, or services. The problem of sparse vector representation exists in handling the large-scale data and also semantics of words are not considered in many existing works. Effective word embedding techniques improve the task of sentiment analysis by overcoming the above two problems. Especially to rate a movie, the keywords need to be identified from a movie review, so an attention mechanism is introduced. It is a challenging task, when the review is expressed in multiple sentences. In this scenario, the entire sentence needs to be considered instead of individual words to determine the sentiment from multiple sentences. To achieve this task, we have proposed a novel long short term memory (LSTM)-based deep learning architecture that applies a sentence embedding using the universal sentence encoder along with an attention layer. To evaluate the proposed approach, we have carried out various experiments using both word-embedding and sentence-embedding with LSTM architecture on benchmark collection IMDB dataset. From the experimental results, it is observed that the proposed sentence-embedding with self-attention LSTM method is significantly better than the other word-embedding based approaches, with an improvement of 5%, and we have achieved an F1 score of 89.12% for our approach.
Keywords: sentiment analysis; universal sentence encoder; long short term memory; word embedding; GloVe; self-attention.
Real-time robust tracking with part-based and spatio-temporal context
by Yanxia Wei, Zhen Jiang, Junfeng Xiao, Xinli Xu
Abstract: Owing to the significant and excellent performance of correlation filters in the aspect of computation convenience, correlation filter-based trackers have become increasingly popular in the visual object tracking community. However, complete or partial occlusion is one of the major factors that seriously impact the tracking performance in visual tracking. To address this issue, we propose a novel tracking algorithm that perfectly integrates the results from the global correlation and local correlation filters for estimating the more accurate position of target. Then, we introduce the occlusion detection mechanism to eliminate the occlusion impact on the final position of object. In addition, our proposed tracker employs the spatial geometric constraints among the global object and local patches of object for preserving the structure integration of object. For verifying our method, we conduct extensive qualitative and quantitative experiments on challenging benchmark image sequences.
Keywords: tracking; correlation filter; occlusion; part-based strategy; spatial geometric constraint.
Maximum power production operation of doubly fed induction generator wind turbine using adaptive neural network and conventional controllers
by Hazem Hassan Ali, Ghada Saeed Elbasuony, Nashwa Ahmad Kamal
Abstract: Production of maximum power based on control of the Rotor Side Converter (RSC) of Doubly Fed Induction Generator (DFIG) wind turbine direct axis current is necessary in order to accomplish fast reaching to the maximum power point and protect the working parts in RSC from high current overshoot. An assessment study between adaptive Neural Network (NN) and conventional Proportional Integral (PI) controllers for control of RSC direct axis current is introduced in this paper. NN controller based LevenbergMarquardt backpropagation (LMBP) is designed for its training to mainly control RSC direct axis current. Also, RSC direct axis current is estimated through using PI controller that is used to control the speed of DFIG according to optimum tip speed ratio obtained by genetic algorithm. The simulation results demonstrated that RSC based NN controller are better than RSC based conventional speed regulator in protecting RSC parts from high current overshoot.
Keywords: DFIG wind turbine; MPPT; optimisation; NN controller; conventional controller.
A hierarchical network embedding method based on network partitioning
by Xiankun Zhang, Yunbin Ma, Xuexiong Luo, Junlong Kang
Abstract: In order to analyse the real-world information network more effectively, we propose a hierarchical network embedding method based on network partitioning, NPHNE. NPHNE is a nested algorithm that it can be combined with existing baseline algorithms to enhance their representation. NPHNE consists of two parts: graph abstracting and embedding propagation. The process of NPHNE is as follows: firstly, modularity is used to pre-determine the network partition, the purpose is to constrain the maximum number of levels. Then, based on the hybrid collapsing method, a series of abstract graphs with successively smaller scales are constructed. The representation of the coarsest abstract graph is learned by the existing baseline algorithms. Finally, the representation is propagated and refined level by level from the coarsest abstract graph to the original graph. We evaluate NPHNE on multi-label classification tasks on citation networks and social networks. The experimental results demonstrate that the maximum performance gains over the baseline algorithm by up to 29.1% Macro-F1 score.
Keywords: hierarchical network embedding; network partitioning; modularity; multi-label classification.
What are students thinking and feeling? Understanding them from social data mining
by Hua Zhao, Yang Zuo, Chunming Xu, Hengzhong Li
Abstract: Students' digital footprints on social media shed light into their personal experiences. Mining the results of these social data is useful for educators to understand students' mood swings and provide corresponding helps. But owing to the sharp increase of social data, analysing these data manually is impossible. In this paper, we focus on Chinese college students, and explore a method to better understand them based on social data mining. The method firstly collects the social data related to students, creates a hierarchy category system based on data contents analysis; secondly, it proposes a simple but effective multi-class classification method to classify the data into several appropriate categories; finally, it carries out the sentiment analysis of each one, and then looks deep into their emotion evolutionary process. Experiment results show that postgraduate entrance exam, final exam and other professional certificate exams are three prominent concerns of students, and they express worry about them.
Keywords: social data mining; classification; sentiment analysis; education.
Analysis of web browser for digital forensics investigation
by AlOwaimer AlOwaimer, Shailendra Mishra
Abstract: In today's digitalised world a lot of information is getting online, the size of online data is getting huge day by day and thus emerges the field of data science. The gateway for surfing on the internet is a web browser when there is so much huge size of data it also makes it vulnerable to the people have malicious intentions. Security is being compromised and thus makes the data vulnerable. A browser can be exploited by an internal malicious user and the main hindrance comes when all the browsing data is deleted. A forensics investigation needs to extract all the pieces of evidence, such as like history, cookies, URL, sessions and saved passwords from the cloud space provided by the browser. The research method is a mix of qualitative and quantitative. A scenario is created in a virtual environment in which there is a victim machine whose browser is exploited. Dumpzilla and bulk extractor forensics tools were used to capture its history, URLs, cookies, sessions, add-ons, and extensions, and SQLite database is used in auxiliary with these tools to get the required information. The extracted information is analysed to find the malicious user. Two different platforms are used for the authentication and verification of pieces of evidence collected. The culprit is caught based on matching web browsing activities from the victim machine and another machine in the same place as the victim machine.
Keywords: digital forensics; web browser forensics; forensic investigation; digital evidence.
IoT data security with DNA-genetic algorithm using blockchain technology
by Sultan Saad Alshamrani, Amjath Basha
Abstract: The Internet of Things (IoT) is an on-demand technology that is used in different applications and also includes different sensors, embedded devices and some other objects related to the internet. IoT devices are mainly designed for gathering diverse kinds of data from different sources and transferring data in digitalised form. Moreover, data security is an important issue in IoT technology, which affects the privacy of data. The main objective of this to handle large amounts of IoT data with cryptographic technique for transmission of data with high security, to enhance anonymity, security, and reliability of IoT, to establish consensus and trust over decentralised networks by addressing difficulties in trustless environment, to enable agreement between unique users by using a consensus mechanism in blockchain in a decentralised manner. A new lightweight encryption technique called DNA-GA (deoxyribonucleic acid genetic algorithm) has been proposed for resource-restricted IoT devices which gives high level sensing of data security. Blockchain is another method proposed in this research to store data in the form of blocks using a hash function where the respective user alone can decrypt the data with the gedecomposinerated hash key, so that intruders are not able to hack the original data. Several experiments were conducted with different kinds of sensed data for our proposed method, and the result is compared with the existing method. The results show that the proposed method outperforms the existing method in handling a large amount of IoT data by reducing encryption and decryption time and provide a high level of security to protect the data.
Keywords: DNA; GA; blockchain; IoT; hash function; decentralised network.
An analytical review of texture feature extraction approaches
by Mohammad Reza Keyvanpour, Shokofeh Vahidian, Zahra Mirzakhani
Abstract: Image registration has been an essential task in computer vision and image processing. There are many applications of image registration, for instance, in medical systems. Image registration consists of four main steps, including features extraction, feature matching, transform model estimation, and resampling and transformation. The feature extraction step makes the image registration process more accurate. Despite a large number of survey articles on texture feature extraction approaches, a comprehensive classification of approaches is still required, which also identifies the strengths and weaknesses of each approach. Therefore, the novelty of this paper is that our analytical framework includes three major components: a complete classification of texture feature extraction approaches, using crucial evaluation criteria to present an analytical and qualitative comparison between each approach, which simplifies the accurate selection of the proposed approaches for the intended application. Our framework can also lead to the development of texture feature extraction approaches in future research of scientists.
Keywords: image registration; feature extraction approaches; challenges; benefits; analytical framework.
Data science based landscape ecology for traditional village landscape protection
by Tong Liu, Shijun Wang, Zi Wang, Bingxin Li, Sulin Guo, Baowei Wei
Abstract: With the revitalisation of rural areas and the advancement of new rural construction, at present, the traditional village protection has received more attention. The optimisation of village layout is an essential issue in the protection of traditional villages, which is not only an objective demand for rural development, but also an important way to protect and avoid damage to the ecological and landscape patterns of the original villages. With the increase in the development of traditional villages, the general scale, shape of traditional villages have found profound changes. The phenomenon of village landscape convergence and original landscape destruction occurs from time to time. In the new situation, the traditional village protection faces new opportunities and challenges. By using landscape ecology theory to guide the preservation of traditional villages, the integrity, authenticity and continuity of traditional villages can be maintained. This article takes Yueqing Town, Bailong Village and Shixian Town, Shuinan Village in Yanbian Prefecture, Jilin Province, as the research objects, and applies the landscape pattern index to quantitatively analyse its overall landscape. On this basis, the village landscape optimisation and preservation strategy is proposed, which has certain reference value for the preservation and reuse of traditional villages in Jilin Province.
Keywords: landscape ecology; traditional village protection; landscape protection; landscape pattern index.
A new model of transformer operation state evaluation based on analytic hierarchy process and association rule mining
by Zhenyu Zhou, Haifeng Ye, Huigang Xu
Abstract: In order to establish a transformer state evaluation model for power grid operation and maintenance management, based on the ageing mechanism of the transformer, a state risk evaluation method based on the analytic hierarchy process (AHP) and association rule mining is proposed. Based on the big data analysis of the actual state quantity of the transformer, the subjective weight coefficient of different state quantities is determined by the AHP. The objective weight coefficient of the comprehensive state quantity is determined by the association rules mining. The fusion of the subjective weight coefficient and the objective weight coefficient is completed according to the weight coefficient fusion technology of the mean squares deviation method. The practical results show that the model in this paper can evaluate the operation state of transformer comprehensively and accurately.
Keywords: transformer; ageing mechanism; state evaluation model; analytic hierarchy process; association rules mining.
Comprehensive survey of user behaviour analysis on social networking sites
by Pramod Bide, Sudhir Dhage
Abstract: Social networking sites play an important role in every persons life. Users start expressing their emotions online whenever any humanitarian or crisis-like event occurs. A lot of sub-events are stirred up and the internet gets flooded with people tweeting/posting their opinions. Identifying user behaviours, their content and their interaction with others can help in event prediction, cross-event detection, user preferences, etc. For these reasons, our research process was divided into studying user behaviour with respect to content-centric and probabilistic approaches and a hybrid incorporating the two. We further investigate the existence of multiple OSNs and how they affect user behaviour. The purpose of this paper is to investigate the existing research methodologies and techniques along with discussion and comparative studies. User behaviour analysis is carried out based on content centric, probabilistic and hybrid approach. Content centric analysis dealt with analysis of the content posted which gives rise to various applications, such as gender prediction, malicious users, real-time user preferences, emotional content influence on users, etc. It is observed that in the probabilistic approach, most of the papers addressed employed clustering mechanisms followed by probability distribution for the analysis of user behaviour.
Keywords: social media; user behaviour; content centric features; probabilistic features; hybrid features.
Moth optimisation algorithm with local search for the permutation flow-shop scheduling problem
by Anmar Abuhamdah, Malek Alzaqebah, Sana Jawarneh, Ahmad Althunibat, Mustafa Banikhalaf
Abstract: This work investigates the use of the Moth-Flame Optimisation (MFO) algorithm in solving the permutation flow-shop scheduling problem and proposes further optimisations. MFO is a population-based approach that simulates the behaviour of real moths by exploiting the search space randomly without employing any local searches that may stick in local optima. Therefore, we propose a Hybrid Moth Optimisation Algorithm (HMOA) that is embedded within a local search to better exploit the search space. HMOA entails employing three search procedures to intensify and diversify the search space in order to prevent the algorithm from becoming trapped in local optima. Furthermore, HMOA adaptively selects the search procedure based on improvement ranks. In order to evaluate the performances of MFO and HMOA, we perform a comparison against other approaches drawn from the literature. Experimental results demonstrate that HMOA is able to produce better-quality solutions and outperforms many other approaches on the Taillard benchmark, which is used as a test domain.
Keywords: flow-shop scheduling; flow-shop scheduling problem; makespan; moth-flame optimisation algorithm; local search; adaptive moth optimisation algorithm.
Detection and mitigation of attacks in SDN-based IoT network using SVM
by Shailendra Mishra
Abstract: Adapting software defined networking (SDN) raises many challenges, including scalability and security on the internet of things (IoT) network. The centralised SDN controller in an IoT-SDN network is responsible for managing the critical network operations. Growing network size increases the network load in the controller and faces security challenges, such as cascade failure of controllers, unauthorised access to the controllers, configuration issues, and distributed denial of service (DDoS) attacks. The DDoS attack is one of the most acute threats in the present scenario, and the attacker can exploit the vulnerabilities that are located mostly in the control plane. In previous research studies, authors have found some strategies and proposed some solutions. The attack scenario and security of multiple controller networks are simulated and evaluated in this research. Simulation has been conducted in Mininet-SDN emulator, the hosting OS was Ubuntu Linux, Wireshark was used for analysing the network traffic, and support vector machines were used to classify the traffic flows. DDoS attacks were detected, and mitigation has been done using a support vector machine learning-based approach. The results show that the support vector machine's sensitivity, specificity, and accuracy are excellent in the range of 98.7% to 98.8%. Security solutions are fast and effective in mitigating DDoS attacks.
Keywords: software defined networking; distributed denial of service attack; support vector machine; DDoS mitigation.
A neural adaptive level set method for wildland forest fire tracking
by Aymen Mouelhi, Moez Bouchouicha, Mounir Sayadi, Eric Moreau
Abstract: Tracking of smoke and fire in videos can provide helpful regional measures to evaluate precisely damages caused by fires. In security applications, real-time video segmentation of both fire and smoke regions represents a crucial operation to avoid disaster. In this paper, we propose a robust tracking method for fire regions in forest wildfire videos using neural pixel classification approach combined with a nonlinear adaptive level set method based on the Bayesian rule. Firstly, an estimation function is built with chromatic and statistical features using linear discriminant analysis and a trained multilayer neural network in order to get a preliminary fire localisation in each frame. This function is used to compute an initial curve and the level set evolution parameters, thus providing fast refined fire segmentation in each processed frame. The experimental results of the proposed method prove its accuracy and robustness when tested on different varieties of wildfire-smoke scenarios.
Keywords: fire detection; linear discriminant analysis; neural networks; active contour; level set; Bayesian criterion.
Robust tracking of moving hand in coloured video acquired through a simple camera
by Richa Golash, Yogendra Kumar Jain
Abstract: Interaction of a human with a machine using dynamic hand gestures has become one of the interesting yet challenging areas in many aspects. A hand is a non-rigid, subtle object that moves with varying speed and in an undefined path. Additionally, real-time backgrounds are not stable. RGB data of a moving hand are sensitive to light variation, camera-view, and randomness in behaviour, thus continuous detection and localisation of the hand region in RGB images is strenuous. Some researchers prefer advanced cameras to avoid the mentioned problems and some apply deep learning in their techniques. The first solution puts limitations on the environment and increases the cost of application. The second solution requires a large database for initial training of the deep neural network architecture. In this paper, we provide a unique solution that combines the benefits of scale-invariant feature transform (SIFT) features with automatic feature extraction mechanism of region-based convolutional neural network (R-CNN), a deep learning network, for robust tracking of a moving hand in coloured video acquired through a camera that does not have very high resolution. The efficiency of the proposed methodology is 96.84% in a simple background and 94.73% in a complex background. The comparative analysis of the proposed system with contemporary techniques using RGB images shows that initial hand detection using R-CNN and then tracking using SIFT is capable of tracking hand movement with high accuracy in unconditional background. In the future, the method can be implemented to design user-friendly and economical natural user interfaces.
Keywords: computer vision; deep learning; R-CNN; visual object recognition; feature extraction; scale-invariant feature transform; visual object tracking.
A service-based software architecture for enabling the storage of electronic health records using blockchain
by Ítalo Lima, André Araujo, Rychard Souza, Henrique Couto, Valéria Times
Abstract: The healthcare sector requires computational solutions with reliable authenticity features for the storage and retrieval of electronic health record data. To address this important issue, this article proposes a service-based software architecture to extract data from different legacy databases, standardise the patients clinical data requirements, and store data using different blockchain technologies. To achieve this, a software architecture has been designed to guarantee the independence of data storage technology. In addition, a data meta-schema and a set of mapping rules have been specified to store and organise clinical data following an international health standard. To validate the solution presented, the real world scenario of a Brazilian healthcare institution has been used to evaluate the data extraction, standardisation and storage capabilities in two blockchain platforms widely used in the information technology market.
Keywords: health information systems; blockchain; software architecture; electronic health record.
AI augmentation in the field of digital image processing
by Rahul Malik
Abstract: All of us are very well aware that identifying information visually is more effective and efficient for human intelligence. In this aspect, one of the primary modes of interacting within and around for understanding the situation through images can act as a crucial source of information for the activities related to human intelligence. By identifying these reasons, one can understand the importance of image processing growing progressively. The fast pace in improving technology, particularly in the computer field, creates a base for image processing related applications. This papers primary focus is to accomplish a better image processing impact using AI as part of image processing. The technology related to segmenting the image deals with the division of an image into different regions to extract and identify the features in it. Such problems can be considered as related to combinatorial optimisation. Initially, this paper deals with the introduction, elaboration, and mathematical representation of the essential theory of the ant colony algorithm. Furthermore, part of this paper deals with improvising the global search by introducing the crowding function of fish into the algorithm. At the end of this paper, the improvised algorithm is used to segment the images to improve the impact of the segmentation
process. Using this algorithm, the results demonstrate its feasibility, significant
improvement in performance, and optimisation while segmenting the images.
Keywords: computer vision; image processing; ant colony algorithm; digital image; image segmentation; artificial intelligence algorithm.
Low-cost thermal explorer robot using a hybrid neural networks and intelligent bug algorithm model
by Willian Baunier De Melo, David Calhau Jorge, Vinicius Abrão Marques
Abstract: Autonomous navigation requires an artificial agent able to independently move adapted to the environment. The robot sensors analyse the surrounding environment and learn, from successful exploration experiences, to plot the best routes and avoid obstacles. This paper proposes a new algorithm for navigation being an adaptation of the intelligent bug algorithm (IBA) combined with artificial neural networks. In addition, this approach also aims to reduce costs using low-cost sensors and a proposed thermal measurement system composed of a matrix infrared sensor superposed with a regular camera. The experimental results show that the novel algorithm is efficient, the prototype avoids collisions and manages to optimise the route and the thermal camera demonstrates accuracy in measuring temperatures and identifying different thermal zones. Moreover, the robot's cost reduction and simple operation characteristics make possible its use in destructive missions in a totally inhospitable location for humans, facilitating its implementation for research and testing.
Keywords: explorer robot; raspberry pi; neural networks; autonomous navigation; cost reduction.
A distributed design of ripple-spreading algorithms for path optimisation problems
by Tian-Qi Wang, Gong-Peng Zhang, Xiao-Bing Hu, Hongji Yang
Abstract: As a relatively new nature-inspired algorithm, ripple-spreading algorithm (RSA) exhibits some advantageous features when resolving various path optimisation problems (POPs) against both traditional deterministic algorithms and evolutionary approaches, e.g., RSA is a multi-agent, bottom-up, simulation model full of flexibility for modifications (like many evolutionary approaches), and it can guarantee optimality (like many deterministic algorithms). Towards real applications to large-scale POPs, RSA still needs to improve its computational efficiency. We used to take some measures in order to achieve a tradeoff between computational efficiency and optimality. This paper, for the first time without sacrificing optimality, proposes a way to significantly improve the computational efficiency of RSA. This is done by taking advantage of the multi-agent nature of RSA, i.e., a multi-agent model is naturally friendly to distributed design and parallel computing. Therefore, this paper reports a distributed design of RSA for POPs. Preliminary experimental results clearly demonstrate the effectiveness and efficiency of the new design of RSA.
Keywords: ripple-spreading algorithm; distributed design; parallel computation; path optimisation; computational efficiency.
A less computational complexity clustering algorithm based on dynamic K-means for increasing lifetime of wireless sensor networks
by Anupam Choudhary, Sapna Jain, Abhishek Badholia, Anurag Sharma, Brijesh Patel
Abstract: Clustering in wireless sensor networks is a critical issue based on network lifetime, energy efficiency, connectivity and scalability. Sensor nodes are capable to collect data from any geographical region using routing protocol. This research endeavours to design a less complex computational time clustering algorithm for hierarchical homogeneous wireless sensor networks to extend network lifetime. It forms an optimal number of clusters and reduces the data communication span of sensor nodes using dynamic K-means algorithm. Selection of a suitable cluster head is based on the ratio of the remaining energy of the sensor node to its distance from the centre of the cluster. The simulation results prove that algorithm that has been presented achieves better energy efficiency when compared with other hierarchical homogeneous cluster-based algorithms. It increases the network lifetime, the number of alive nodes per round, the data delivered to the base station, the time of the first node, middle node and last node to die for scalable situations in terms of node density and size of the sensing region.
Keywords: wireless sensor network; sensor node; hierarchical homogeneous cluster-based protocols; cluster Head; base station; network lifetime.
Reduced-order modelling of parameterised incompressible and compressible unsteady flow problems using deep neural networks
by Oliviu Sugar-Gabor
Abstract: A non-intrusive reduced-order model for nonlinear parametric flow problems is developed. It is based on extracting a reduced-order basis from full-order snapshots via proper orthogonal decomposition and using both deep and shallow neural network architectures to learn the reduced-order coefficients variation in time and over the parameter space. Even though the focus of the paper lies in approximating flow problems of engineering interest, the methodology is generic and can be used for the order reduction of arbitrary time-dependent parametric systems. Since it is non-intrusive, it is independent of the full-order computational method and can be used together with black-box commercial solvers. An adaptive sampling strategy is proposed to increase the quality of the neural network predictions while minimising the required number of parameter samples. Numerical studies are presented for unsteady incompressible laminar flow around a circular cylinder, transonic inviscid flow around a pitching NACA0012 aerofoil and a gust response for a modified NACA0012 in subsonic compressible flow. Results show that the proposed methodology can be used as a predictive tool for unsteady parameter-dependent flow problems.
Keywords: non-intrusive parameterised reduced-order model; artificial neural networks; proper orthogonal decomposition; incompressible and compressible flow model order reduction.
Fusion-based Gaussian mixture model for background subtraction from videos
by T. Subetha, S. Chitrakala, M. Uday Theja
Abstract: Human Activity Recognition (HAR) aims at realising and interpreting the activities of humans from videos and it comprises background subtraction, feature extraction and classification stages. Among those stages, the background subtraction stage is mandatory to achieve a better recognition rate while analysing the videos. The proposed Fusion-based Gaussian Mixture Model (FGMM) background subtraction algorithm extracts the foreground from videos, which is invariant to illumination, shadows, and the dynamic background. The proposed FGMM algorithm consists of three stages: background detection, colour similarity, and colour distortion calculation. Here, the Jefries-Matusita distance measure is used to check whether the current pixel matches the Gaussian distribution and, by using this value, the background model is updated. Weighted Euclidean-based colour similarity measure is used to eliminate shadows and a colour distortion measure is adopted to handle illumination variations. The extracted foreground is binarised to easily extract the interest points and the foreground, which has white pixel is stored into the frame. This algorithm is tested over test sets gathered from publicly available benchmark datasets, including the Kth dataset, Weizmann dataset, PETS dataset, and change detection dataset. Results have proved that the proposed FGMM exhibits better accuracy in foreground detection, with an increased accuracy compared with the prevailing approaches.
Keywords: human activity recognition; Gaussian mixture model; fusion-based Gaussian mixture model; background subtraction.
Design and analysis of search group algorithm-based PD-PID controller plus redox flow battery for automatic generation control problem
by Ramana Pilla, Tulasichandra Sekhar Gorripotu, Ahmad Taher Azar
Abstract: The ability of a redox flow battery (RFB) is analysed in the present paper to minimise the tie-line power and frequency deviations of the five-area thermal power system. Initially, a power system network with five areas and a nonlinearity of generation rate constraint is designed in MATLAB/SIMULINK environment. After that a proportional derivative-proportional integral derivative (PD-PID) controller is evaluated for the proposed system. Finally, the RFB is installed in area-1, area-2, area-3, area-4 and area-5 for dynamic response enhancement. Results of simulation show that better transient response characteristics can be obtained by using PD-PID controller along with RFB in area-1. The robust analysis is also performed to show the capability of the proposed method.
Keywords: dynamic response; generation rate constraint; PD-PID controller; redox flow battery; search group algorithm; transient response.
Accurate detection of network anomalies within SNMP-MIB dataset using deep learning
by Ghazi Al-Naymat, Hanan Hussain, Mouhammd Al-Kasassbeh, Nidal Al-Dmour
Abstract: An efficient algorithm for supporting the intrusion detection system (IDS) is required for identifying unauthorised access that attempts to collapse the confidentiality, integrity, and availability of computer networks. The machine learning approaches such as (a) multilayer perceptron, (b) support vector machines, (c) nearest neighbour classifiers and (d) ensemble classifiers, such as Random Forest (RF) show higher accuracy only when the additional feature selection techniques such as Infogain, ReliefF, or Genetic Search are used. When the data gathered for training and testing is huge with a greater number of features, the extra computation of feature selection will result in a higher consumption of hardware resources (CPU, memory, and bandwidth). On the other hand, another subset of the machine learning approach called the Deep Learning (DL) algorithm does the feature selection, automatically to overcome this limitation. In this paper, a deep learning method called Stacked Autoencoder (SAE) is proposed for detecting seven different types of network anomaly using the SNMP-MIB dataset. The autoencoder is a variant of the neural network, which transforms the set of n inputs to a different set of m reduced number of outputs (encoding). Previous outputs are then processed by the decoding part to get the desired output of n dimensions, which is identical with the initial input. They are stacked one by one to form a deep SAE. Parameters of the model are selected by trial and error method to get the best training functions, activation functions, learning rate, etc. The proposed deep learning method attains a high accuracy of 100% and saves the extra computations and resources spent on feature selection. The proposed model is also compared with 22 prominent machine learning techniques from the following categories: (i) decision trees, (ii) discriminant analysis, (iii) support vector machines, (iv) nearest neighbour classifiers and (v) ensemble classifiers. It is found that our model outperforms all other machine learning algorithms in terms of accuracy, precision, and recall.
Keywords: deep learning; DoS; network anomalies; SNMP-MIB; detection.
An interpolation algorithm of B-spline curve based on S-curve acceleration/deceleration with interference pre-treatment
by Guirong Wang, Qi Wang
Abstract: A B-spline curve interpolation algorithm based on S-curve acceleration/deceleration (ACC/DEC) with interference pretreatment is proposed to achieve a smooth transition of feed-rate, and to reduce the impact with acceleration mutation in computer numerical control (CNC) machining. According to the demand of chord error, the algorithm can adaptively adjust the feed-rate of each interpolation point, and divide a B-spline curve by velocity cusps. The interference points of the whole curve can be found out by using the S-curve ACC/DEC calculation for the velocity cusps of the whole curve from the forward and reverse directions. Then, the feed-rate of interference points is re-determined to avoid jerk overrun on the curve segments between mutual interference points, so as to improve the processing stability of the machine tool. The simulation and experiment results demonstrate that the algorithm can obtain the smooth transition of feed-rate and acceleration in CNC machining, and ensure that the jerk can meet the ACC/DEC of system. The CNC system can use this method for high-precision and high-speed machining of complex products.
Keywords: S-curve ACC/DEC; interference pre-treatment; piecewise curve; interference points; feed-rate scheduling; CNC machine tools; interpolation algorithm.
A new multistable jerk system with Hopf bifurcations, its electronic circuit simulation and an application to image encryption
by Sundarapandian Vaidyanathan, Irene M. Moroz, Ahmed A. Abd El-Latif, Bassem Abd-El-Atty, Aceng Sambas
Abstract: In this work, we announce a new 3-D jerk system and show that it is chaotic and dissipative with the calculation of the Lyapunov exponents of the system. By performing a detailed bifurcation analysis, we observe that the new jerk system exhibits Hopf bifurcations. It is also shown that the new jerk system exhibits multistability behaviour with two coexisting chaotic attractors. An electronic circuit simulation of the jerk system is built using Multisim. Finally, based on the benefits of our proposed chaotic jerk system, we design a new approach to image encryption as a cryptographic application of our chaotic jerk system. The simulation outcomes prove the efficiency of the proposed encryption scheme with high security.
Keywords: bifurcations; chaos; chaotic systems; circuit design; jerk systems; image encryption.
Compensation of variability using median and i-Vector+PLDA for speaker identification of whispering sound
by Vijay Sardar
Abstract: Speaker identification from the whispered voice is troublesome contrasted with neutral as the voiced phonations are missing in the whisper. The success of the speaker identification system mainly depends on the selection of appropriate audio features. The various available audio features are explored here and it is shown that the timbre features are able to identify the whispering speaker. Only the well-performing, and thus limited timbre, features are sorted by the hybrid selection algorithm. The timbre features named brightness, roughness, roll-off, MFCC and irregularity using CHAIN database offer improvement in the identification outcomes by 5.8% over the baseline system. The framework ought to be robust enough to repay intra-speaker and inter-speaker variability, including channel impacts. The analysis using timbre features based on median value predicted that the intra-speaker variability is being compensated. The use of median timbre features reported further enhancement of 1.12% compared with using timbre features and a further decline in False Negative Rate (FNR). The use of i-Vector + probabilistic discriminant analysis (PLDA) and Support Vector Machine (SVM - cosine kernel) have contributed a relative improvement in accuracy of 8.13%. The reductions in False Positive Rate (FPR) and False Negative Rate (FNR) confirm better variability compensation.
Keywords: whispered speaker; median timbre feature; i-Vector; cosine kernel; support vector machine.
A Model Predictive Control Strategy for Field-circuit Coupled Model of PMSM
by zhiyan zhang, pengyao guo, yan liu, hang shi, Yingjie Zhu , Hua Liu
Abstract: Based on the analysis of the mathematical equations and drive circuit of permanent magnet synchronous motor (PMSM), a model predictive control strategy for the controller of PMSM is proposed. The stator current discretization model and the cost function of model predictive control are established, and voltage vector selection is derived. Then, the coupling mechanism among motor, driver and controller is analyzed?and the field-circuit coupled model of a 1kW PMSM using model predictive control is set up. Next, the starting performance, load characteristics and electromagnetic field of the motor are obtained. Good speed and electromagnetic characteristics verify the effectiveness of the PMSM control strategy and the correctness of the PMSM field-circuit coupled model. Finally, the back EMF waveforms and its harmonics of the field-circuit coupled model and the finite-element model without drive circuit and controller are compared and analyzed. The simulation results show that the amplitude of back EMF in both models is basically the same while the field-circuit coupled model has high THD value, which can simulate the practical conditions.
Keywords: PMSM, model predictive control, voltage vector selection, field-circuit coupled model
Mechanics of the tubing string for supercritical CO2 fracturing
by Wenguang Duan, Baojiang Sun, Deng Pan, Hui Cai
Abstract: Supercritical CO2 fracturing is one of the most efficient ways for increasing petroleum productivity. The tubing string for the fracturing is necessary and plays an in important role in the fracturing process. A mechanical model of the tubing string in the well for fracturing is set up. The forces on the tubing string are analysed. The mechanical formulas are derived. The stresses on the tubing string are calculated and the strength of the tubing string is checked. The running accessibility of the tubing string through the well for fracturing is studied. The equations for calculating the critical force on the tubing string for fracturing causing sinusoidal bucking and helical bucking are given. Based on the finite element method, a model is set up. The stress and deformation of the tubing string in the horizontal and deviation well sections are calculated. Results show that under the given conditions, the tubing string is safe and efficient.
Keywords: supercritical CO2; fracturing; tubing string; running accessibility; mechanics.
Ontology-based broker system for interoperability of federated cloud computing platforms
by Surachai Huapai, Unnadathorn Moonpen, Thepparit Banditwattanawong
Abstract: This paper presents an ontology-based broker system for the interoperability of federated clouds. The system can provision cloud infrastructure resources from different platforms to meet the users requirements of Infrastructure as a Service (IaaS). The system engaged an ontology to enable the interoperability of heterogeneous IaaS management platforms, OpenStack, Apache CloudStack, and VMware ESXi. The system provisioned appropriate cloud-infrastructure resources from available platforms based on a vector-space algorithm. Evaluation results relying on the two datasets of non-scheduled and scheduled IaaS-user requests show that our system is practical in that average latencies to generate REST commands for virtual machine provisioning take less than a second per request and are linearly proportional to the number of provisioned servers.
Keywords: federated-cloud computing; cloud broker; cloud ontology; infrastructure as a service; interoperability.
On the estimation of makespan in runtime systems of enterprise application integration platforms: a mathematical modelling approach
by Fernando Parahyba, Rafael Z. Frantz, Fabricia Roos-Frantz
Abstract: Integration platforms are tools developed to support the modelling, implementation and execution of the integration processes, so that data and functionality from applications in software ecosystems can be reused. The runtime system is a key piece of software in an integration platform and it is directly related to its performance; and, makespan is a metric used to measure performance in this systems. In this paper we propose a mathematical model to estimate the makespan for integration processes that run on application integration platforms built based on the theoretical task-based model. Our model has shown to be accurate and viable to assist software engineers in the configuration and deployment process of integration processes on an actual integration platform. The model was validated by means of a set of experiments, which we report in the paper.
Keywords: enterprise application integration; makespan; runtime system; mathematical modelling; integration platforms.
Gradient iterative based kernel method for exponential autoregressive models
by Jianwei Lu
Abstract: Two kernel method based gradient iterative algorithms are proposed for exponential autoregressive (ExpAR) models in this study. A polynomial kernel function is utilized to transform the ExpAR model into a linear-parameter model. Since the order of the linear-parameter model is large, a momentum stochastic gradient algorithm and an adaptive step-length gradient iterative algorithm are developed. Both these two algorithms can estimate the parameters with less computational efforts. Finally, a simulation example shows that the proposed algorithms are effective.
Keywords: ExpAR model; kernel method; linear-parameter model; momentum stochastic gradient algorithm; adaptive step-length; gradient iterative algorithm.
A Novel Chaotic Grey Wolf Optimization for High-Dimensional and Numerical Optimization
by Mengjian Zhang, Daoyin Long, Dandan Li, Xiao Wang, Tao Qin, Jing Yang
Abstract: Aiming at the weakness of the current evolutionary algorithms for high-dimensional and numerical optimization problems of global convergence, a novel chaotic grey wolf optimization (NCGWO) is proposed for solving the high-dimensional optimization problems. Firstly, the six chaotic one-dimensional maps are introduced and their mathematical models are improved with their mapping ranges being in the interval (0, 1). Secondly, the diversity experiments are conducted to test the results of the chaotic maps. The experiments show that the initial population by chaotic maps is superior to the GWO algorithm and the Sine map is best. Finally, the CSGWO algorithm is also proposed based on the NCGWO algorithm with the parameter C by Sine map. The simulations demonstrate that the performance of the GWO algorithm can be improved by the chaotic maps for high-dimensional and numerical optimization problems, and the effectiveness of the CSGWO algorithm is superior to other evolutionary algorithms and achieves better accuracy and convergence speed.
Keywords: Chaotic system, grey wolf optimization (GWO), chaos initialization, optimization, High-dimension
Recursive identification of state space systems with colored process noise and measurement noise
by Fang Zhu, Xuehai Wang
Abstract: This paper concerns the modeling and identification of the state space system, in which both colored process noise and measurement noise are encountered. By using the state filtering, the state space system with colored process noise is transformed into a model without correlated noise, and a state filtering based parameter estimation algorithm is derived on the base of designing a state filter observer using the multi-innovation identification. The validity of the proposed algorithm is verified by given simulation examples
Keywords: Parameter estimation; Recursive identification; Filtering technique; State estimation.
Research on Cable Partial Discharge Detection and location system based on optical fiber timing
by Jian-jun Zhang, Fang Peng, An-ming Xie, Yang Fei
Abstract: Partial discharge (PD) is an important index to reflect the running state of cable. According to the characteristics and propagation mechanism of cable partial discharge signal, a cable partial discharge detection and location system based on optical fiber time synchronization technology and traveling wave double terminal location principle is developed. The system has high detection sensitivity, high reliability, real-time detection, diagnosis and positioning of cable discharge power supply. The experimental results show that the positioning accuracy of the cable partial discharge source can be effectively improved based on the fiber timing and double terminal positioning technology, and the positioning accuracy can reach 1%; the method studied in this paper can meet the requirements of the accurate location of the partial discharge source of the cable, Gil and other equipment.
Keywords: Cable, partial discharge, optical fiber timing, double terminal positioning, online monitoring.
Life-threatening arrhythmias recognition by pulse-to-pulse intervals analysis
by Lijuan Chou, Yongxin Chou, Jicheng Liu, Shengrong Gong, Kejia Zhang
Abstract: Tachycardia, bradycardia, ventricular flutter and ventricular tachycardia are the four life-threatening arrhythmias, which are seriously harmful to the cardiovascular system. Therefore, a method for identifying these arrhythmias by pulse-to-pulse intervals analysis is proposed in this study. First, the noise and interference are wiped out from the raw pulse signal, and the clear pulse signal is spitted into pulse waves by pulse troughs whose first-order difference are the pulse-to-pulse intervals. Then, fifteen features are extracted from the pulse-to-pulse intervals, and the two-samples Kolmogorov-Smirnov test is utilized to select the markedly changed features. Finally, we design the classifiers for arrhythmias recognition by the probabilistic neural network (PNN), feedback neural network (BPNN) and random forest (RF). The pulse signal from the international physiological database (PhysioNET) is utilized as the experimental data. The experimental results show that RF classifier has the best average classification performance with the kappa coefficient (KC) of 98.86±0.13% which is higher than that of BPNN with KC of 70.70 ± 1.61% and PNN with KC of 62.07 ± 0.75%. Compared with the existing methods, the proposed method has a higher performance in recognition of the four arrhythmias and has great potential to monitor the life-threatening arrhythmia in m-health.
Keywords: Pulse signal, pulse-to-pulse intervals, lifer-threatening arrhythmias, intelligent recognition.
Special Issue on: Advanced Big Data and Artificial Intelligence Technologies for Edge Computing
Efficient synergetic filtering in big datasets using neural network technique
by B. Mukunthan
Abstract: Presently, great accomplishment on speech-recognition, computer-vision and natural-language processing has been achieved by deep neural networks. To tackle the major trouble in synergetic- or collaborative-filtering is the idea of hidden feedback. In this task we concentrated intensively on the techniques based on neural networks. Although a few recent researchers have employed deep learning, they mostly used it to sculpt auxiliary facts, along with textual metaphors of objects and acoustic capabilities of musics. When it involves the major aspect in synergetic filtering, the communication between customer and object capabilities, they still resorted to matrix factorisation and implemented a core product on the hidden capabilities of customers and objects. We present a popular framework named Artificial Neural Synergetic Filtering (ANSF) to substitute the core makeup with a neural design which could be very efficient to analyse data with a random feature. ANSF is ordinary and can specifically popularise matrix-factorisation beneath its framework. To improvise ANSF modelling with non-linearities we propose to leverage a multi-layer perceptron to investigate customer-object communication functions. In-depth experiments on actual-global databases display big improvement of our proposed ANSF over the latest techniques. Investigational results show that the application of core layers of artificial neural networks gives improved overall performance.
Keywords: synergetic filtering; big data; matrix factorisation; deep neural network; multi-layer perceptron.
Computer image analysis for various shading factors segmentation in forest canopy using convolutional neural networks
by Liangkuan Zhu, Jingyu Wang, Kexin Li
Abstract: Determination of images of the various parts in the forest canopy is critical because it reflects a variety of parameters for plant population growth in forest ecosystems. This study presents the use of deep learning for the detection of various shading factors in hemispherical photographs of the forest canopy. First, a forest canopy hemispherical photographs dataset that can be used for the research of related algorithms is constructed. Based on FCN, the dilated convolution layers and multi-scale feature fusion are used to improve the accuracy of forest canopy image segmentation. Furthermore, the Conditional Random Field (CRF) is adopted to optimise the results. Finally, experiments show that this method can achieve automatic segmentation of the sky, leaves, and trunks of forest canopy images. Compared with the FCN model, the average pixel accuracy of the improved FCN model is improved by 9.11%, and it has good robustness.
Keywords: hemispherical photographs; image segmentation; full convolutional neural network; dilated convolution; multi-feature fusion; conditional random field.
Special Issue on: Computational Advances in Healthcare Solutions
Unification of firefly algorithm with density-based spatial clustering for segmentation of medical images
by Bandana Bali, Brij Mohan Singh
Abstract: This paper proposes a computer-aided approach for brain image segmentation to figure out various characteristics of digital images that are responsible for the identification of brain tumours with MRI images. The proposed Density-Based Spatial Clustering Fused with Firefly (DB-FF) method is based on density-based spatial clustering and firefly algorithm, which have significant places in nature-inspired computing techniques. In this research, the solutions of the firefly algorithm have been improved by the density-based spatial clustering algorithm, and a soft computing criterion has also been used as a fitness function. The proposed method has been tested on commonly used images from Harvard Whole Brain Atlas, and the results of this method have been compared with other standard benchmarks from the survey. The proposed DB-FF method achieved better segmentation than standard segmentation quality metrics, such as normalised peak signal to noise, normalised root square mean error and structural similarity index metric. Matlab has been used for implementation and observation. The result demonstrates that the proposed method has a better and more robust performance as compared with the existing MRI segmentation models.
Keywords: brain tumour detection; data clustering technique; firefly algorithm; image segmentation.
Alzheimer's disease diagnosis based on feature extraction using optimised crow search algorithm and deep learning
by Sonal Bansal, Aditya Rustagi, Anupam Kumar
Abstract: Alzheimers Disease (AD) is a long-lasting, progressive, cognitive disorder of degenerative nature and one of the most common reason to cause the dementia. Dementia leads to decline in thinking capacity, inability to handle the behavioural and social skills, and disrupts the normal functioning of the persons ability. Conventional methods of assessing the symptoms and information from a close family member are being recorded to analyse the effect of the disease and its stages. Neuroimaging is one of the best methods being used by neurologists and doctors for Alzheimers disease. MRI is being used around the world for the diagnosis of the disease and to provide insights to the brain and its functioning. With the advances in the area of machine learning, the application to various medical images such as MRI and CT scan, is on the rise and has become a major discipline of research among the experts and analysts. Existing methods of feature extraction from an image involve CNN, which provides a large number of feature sets that require great computation power as well as time to evaluate them using traditional machine learning or any deep learning algorithm. Consequently, we propose an Optimised Crow Search Algorithm (OCSA) for early diagnosis of AD which, when applied to the raw MRI image features, yields a highly representative dense embedding of the same. The mapping learned between this embedding and the image labels resulted in diagnosing 98.62% of AD patients dataset correctly.
Keywords: Alzheimer’s disease; magnetic resonance images; evolutionary algorithm; feature extraction; intelligent computer-aided diagnosis systems; medical imaging; medical informatics.
An intelligent COVID-19 classification model using optimal grey-level co-occurrence matrix features with extreme learning machine
by Pavan Kumar Paruchuri, V. Gomathy, E. Anna Devi, Shweta Sankhwar, S.K. Lakshmanaprabu
Abstract: In recent times, earlier diagnosis of 2019 novel coronavirus disease (COVID-19) is essential for disease cure and control. Chest computed tomography (CT) images are found to be a reliable, helpful and faster method for the classification of COVID-19. Since the chest CT image diagnosis requires medical experts and more time, an automated intelligent model needs to be developed for effective COVID-19 diagnosis. This paper presents a new automated COVID-19 diagnosis model using optimal grey level co-occurrence matrix (GLCM) based feature extraction and extreme learning machine (ELM) based classification. The input chest images undergo preprocessing to improve the image quality. Next, the optimal GLCM features are derived by the use of Elephant Herd Optimisation (EHO) algorithm. Then, the ELM model is applied to perform the classification task. The performance of the OGLCM-ELM model has been validated using the benchmark dataset and the experimental outcome ensured the superior performance of the proposed model over the compared methods. The proposed OGLCM-ELM model has achieved maximum sensitivity of 89.56%, specificity of 90.45%, F-score of 90.13% and accuracy of 90.69%.
Keywords: COVID-19; disease diagnosis; feature extraction; classification; deep learning.
Med-Net: a novel approach to ECG anomaly detection using LSTM auto encoders
by Koustav Dutta, Rasmita Lenka, Soumya Ranjan Nayak, Asimananda Khandual, Akash Kumar Bhoi
Abstract: Time series data is generated in various sectors of day-to-day life. Among all, the time series data plays vital role in medical domain analysis. In this specific context, various continuous time series dependent EEG and ECG signals are the most important. Till now, heavy reliance on doctors regarding the manual analysis of these signals for understanding, monitoring and detecting the anomaly is cumbersome. Thus, this paper proposes a novel approach to analyse and detect ECG signals for tracking of anomalies using Hybrid Deep Learning Architectures (HDLA). The proposed scheme works by implementing self-supervised pattern recognition according to the mechanism of Long Short Term Memory networks (LSTM) in terms of auto encoder and decoder. Finally, the proposed scheme is tested on Physionet dataset. The outcome of the model can also handle noise associated with ECG-based time series signals, and it achieved accuracy and solved the overfitting problems.
Keywords: bio-signals; encoder; decoder; LSTM; ECG; anomaly; time series; reconstruction error.
Multimodality medical image fusion based on non-subsampled contourlet transform
by Velmurugan Subbiah Parvathy, Sivakumar Pothiraj, Jenyfal Sampson
Abstract: The fusion of medical images is the mainly essential and effective technique for disease analysis. We have provided a Non-Subsampled Contourlet Transform (NSCT) image fusion technique using Neuro Fuzzy with Binary Cuckoo Search (NFBCS) and the Slap Swarm Optimisation (SSO) method. Here we successfully fused the Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) images and created a single merged image, which provides a new integrated diagnostic method. Initially, two unique sets of images, for example, MRI and CT, were considered for the fusion procedure. These pairs of images are used for NSCT to generate the image to divide the high frequency module and the low frequency module. The mixing policies are used here to generate and combine high and low frequencies. Compared with other existing techniques, the results of the proposed technical tests show better processing efficiency and deliver results on subjective and objective evaluation criteria. This is particularly advantageous for accurate clinical analysis.
Keywords: magnetic resonance imaging; computed tomography; neuro fuzzy; binary cuckoo search; slap swarm optimisation.
Efficient detection of supraventricular tachycardia by machine learning techniques
by Monalisa Mohanty, Asit Subudhi, Mihir Narayan Mohanty
Abstract: Supraventricular tachycardia (SVT) refers to abnormally fast heartbeat that arises because of the improper electrical activity in the upper chamber of the heart. Though SVT cases nowadays are taken as a less hazardous disease, however recurrent incidents may degrade the heart muscle over a period of time. Tachycardia usually refers to a quick rise in the heart rate that is of more than 100 beats per minute. SVT is a kind of arrhythmia that is based on an abnormal heartbeat. Electrocardiogram (ECG) is one of the most significant diagnostic tools used for the recognition of the health of a heart. The increasing number of heart patients has led to essential progress in the techniques of automatic detection of the numerous kinds of abnormality or the arrhythmia of the heart to reduce the pressure and share the load of the physicians. ECG recordings have been acquired from MIT-BIH supraventricular arrhythmia database (SVDB) of the Physionet repository. Each record consists of ST, N and VF rhythm with a duration of 30 minutes length. Then using various techniques, a set of features has been extracted for ST,N and VF and finally fed into a classifier, such as logistic model tree or multilayer perceptron, to classify the ECG signals.
Keywords: tachycardia; supraventricular tachycardia; arrhythmia; decision tree classifiers.
Application and evaluation of classification model to detect autistic spectrum disorders in children
by Hrudaya Kumar Tripathy, Pradeep Kumar Mallick, Sushruta Mishra
Abstract: A child affected by Autism Spectrum Disorder (ASD) faces significant difficulties in social interaction (i.e., communication with language, understanding emotional states of others, thinking and behaving together, etc.). So there is a requirement for a real-time and easy-to-access diagnostic model to identify autism during the initial phase of occurrence to assist medical experts. Presently, an efficient cure for autism does not exist. A reliable detection model will help to provide better therapy, thereby supporting autistic children to continue a better life. This work deals with the autistic dataset's efficient categorisation using various classifiers, such as naive Bayes, neural network, and random forest. At the same time, Python is a programming tool to determine algorithms with optimum accuracy by multiple simulations. An autistic dataset from the UCI repository was used for this research study. It is used to build a model where parents of a suspected autistic child can detect autism by providing their answers to some particular questions relating to autism characteristics. The implementation result indicated that the random forest classifier gave the optimal performance. A 97.5% classification accuracy rate was generated using a random forest algorithm. An RMSE value of 0.676 was observed. A very minimal 1.16 sec execution time was recorded using the autistic dataset. The recorded values were 92.8%, 92.6%, 90.8%, and 91.5% for the mean accuracy, precision, recall, and f-score metrics. Thus it can be concluded that autistic disorder detection using random forest classifiers generated an optimal performance.
Keywords: autism syndrome disorder; naive Bayes; neural network; random forest; logistic regression.
IoT implementation strategies amid COVID-19 pandemic
by Dhawan Singh, Aditi Thakur, Maninder Singh, Amanpreet Sandhu
Abstract: The world, at present, is witnessing grave challenges to its established institutions and shared beliefs owing to the outbreak of novel coronavirus. Almost all of our establishments are under threat and unprecedented disruptions are being witnessed across all spheres of life. Besides the medical hunt for discovering the cure, there exists an equally significant need to invent technological solutions for restoring numerous services while considering the restrictions imposed by the pandemic. Therefore, in this research work, we have investigated and analysed the possibilities, opportunities, and applications of IoT technology in the field of food safety and quality control, automatic disinfection, healthcare systems, wearable health devices, and personal hygiene. We have assessed various features of currently available IoT design platforms and standard protocols and proposed feasible and dynamic strategies for their implementations. The efficacy of the system demonstrates an immense possibility for the continuation of IoT-based technology, even after the novel coronavirus scare is over.
Keywords: COVID-19; disinfection; food safety; internet of things; pandemic; smart technology; wearable health devices.
Performance analysis of surrounding cylindrical gate all around nanowire transistor for biomedical application
by Amit Agarwal, Prashanta Chandra Pradhan, Bibhu Prasad Swain
Abstract: This paper proposes a highly sensitive, more accurate, and faster device using silicon on insulator based cylindrical surrounding gate all around nanowire (SCGAA-NW) transistor for biomedical sensor applications and different chemicals present in the blood or environment by setting and analysing proper physical parameters of the device. Different physical parameters, channel material (i.e. SiN, CdS, GaN, ZnO, GaP, Si, GaAs, Ge), oxide material (i.e. SiO2, Si3N4, Al2O3, Er2O3, Y2O3, HfO2, Ta2O5, La2O3), channel radius (1-10 nm), oxide thickness (1-10 nm), the concentration of different material have been employed on the sensor acting as a gate to source voltage, drain to source voltage (−0.5 V to 0.5 V), channel doping (107 to 1014) respectively for the best suitable biomedical application in different environments. Analytical modelling of SCGAA-NW transistor was performed by solving 1-D Poisson's equation, using Gauss law and parabolic approximation method using Matlab simulator.
Keywords: biosensor; bio-medical application; GAA-NW; gate all around nanowire; MOSFET; cylindrical channel; potential at the channel surface.
Fuzzy logic system for diabetic eye morbidity prediction
by Tejas V. Bhatt, Raksha K. Patel, Himal B. Chitara, Gonçalo Marques, Akash Kumar Bhoi
Abstract: Diabetes is a common chronic disease; the number of people who are affected by this health problem is increasing worldwide, leading to a high cost for healthcare systems. Therefore, the main contribution of this paper is to present a fuzzy logic system for diabetic eye morbidity prediction. This work is divided into two parts. The first part is the examination of eye vision by the ophthalmologist and also other examinations such as postprandial blood sugar, hypotension, cholesterol, and duration of diabetes. The second part is the analysis of 400 patients' medical records collected in 2019. The fuzzy system proposed for prediction of diabetes retinopathy provides reliable accuracy for eye vision-threatening and eye morbidity. The proposed fuzzy system has five input parameters and one output parameter, which predicts diabetic neuropathy. The input parameters are random blood sugar, postprandial blood sugar, hypotension, cholesterol, and eye vision. The output parameter is the morbidity in diabetic retinopathies, which are non-proliferative, proliferative, clinically significant macular oedema. The proposed system is designed to support the endocrinologist and ophthalmologist in the diagnosis of diabetic retinopathy.
Keywords: artificial intelligence; blood sugar; cholesterol; diabetic retinopathy; fuzzy set theory.
Survivability prediction of patients suffering hepatocellular carcinoma using diverse classifier ensemble of grafted decision tree
by Ranjit Panigrahi, Moumita Pramanik, Udit Kumar Chakraborty, Akash Kumar Bhoi
Abstract: The mortality rate of patients who have cancer is the second highest cause of death around the globe. Hepatocellular Carcinoma (HCC), a type of liver cancer, is once such a cause of death. Though the probability of survival of patients is very rare, a mechanism to predict chances of survivability will provide a great aid to the medical practitioners to treat patients suffering from HCC. In this article, two state of the art survivability prediction schemes have been proposed separately for male and female subjects suffering HCC. The prediction engine employs Feature Selection Via (FSV) concave minimisation feature ranking and Sigmis feature selection scheme to extract limited features of both male and female subjects and an ensemble of decision tree grafting mechanism successfully predicts the chances of survivability of HCC patients. The gender-specific survivability prediction engine is the first-ever such prediction model for the diagnosis of HCC.
Keywords: hepatocellular carcinoma; HCC; liver cancer detection; Sigmis; FSV; machine learning.
Detection of bifurcations and cross-over points from retinal vasculature map using modified window feature-point detection approach
by Meenu Garg, Sheifali Gupta, Soumya Ranjan Nayak
Abstract: Identification of feature points such as bifurcation points and cross-over points in retinal fundus images is useful for predicting the various cardiovascular diseases. In this paper, a new approach called the Modified Window Feature-point Detection (MWFD) is proposed to identify the vascular feature points in the fundus image. The MWFD technique makes use of two different windows 3×3 and 5×5 with alternative vessel pixel property for the detection of all feature points. This paper also resolves the problem of conversion of one cross-over point into two bifurcation points generated due to skeletonisation. The simulation is performed on the DRIVE database using MATLAB software. Simulation results show quantitative improvements in detection of feature points from retinal fundus image by increasing the number of true positives and reducing false positives and false negatives. These results provide an efficient and reliable technique for analysis of various retinal structures.
Keywords: fundus image; retinal vasculature map; feature point; bifurcation point; cross-over point; retinal detachment.
Decision-making on the existence of soft exudates in diabetic retinopathy
by A. Reyana, V.T. Krishnaprasath, Sandeep Kautish, Ranjit Panigrahi, Mahaboob Shaik
Abstract: Medical image analysis is recognised to be the most important research area for diagnosis and screening of a wide range of medical problems. The commonly discussed diabetic retinopathy has become a vital factor of serious eye diseases leading to eye blindness. Diabetes mellitus, a remarkable metabolic disorder, is rapidly increasing health threats worldwide. The major cause of Diabetic Retinopathy (DR) is the increase in blood insulin levels. The DR lesion thresholds inferred have protective effects and have no benefits for patients. From the fundus images the prediction of micro-aneurysms is still the major challenge. Micro-aneurysms formulation is the initial sign of DR to reduce the risk of non-proliferated DR. With this in mind, there exists a need for a diagnostic system, for early detection of DR to be used by the ophthalmologist to identify different types of lesions like haemorrhages and exudates. This paper presents a new approach to detect and classify exudates in coloured retinal images, eliminating the replication of exudates by removing the optic disc region. Our research aimed to extend the current knowledge of several image processing techniques including image enhancement, segmentation, classification, and registration for early diagnosis preventing visual impairment and blindness.
Keywords: diabetic; retinopathy; homogeneity; segmentation; blindness.
Prediction of abnormal hepatic region using ROI thresholding based segmentation and deep learning based classification
by Shubham Kamlesh Shah, Ruby Mishra, Bhabani Shankar Prasad Mishra, Om Pandey
Abstract: This paper proposes a novel Computer-Aided Diagnosis System (CADS) model using Artificial Intelligence (AI) to segment liver form abdomen CT scan. Deep Learning Convolutional Neural Network (DL-CNN) model is proposed to train the program to classify normal and abnormal liver images. For training data set generation, another novel Region of Interest (ROI) based thresholding image processing technique is proposed. DL-CNN network is also compared with the basic CNN model to understand the difference between basic and deep learning networks. The basic CNN model yielded an accuracy of 50.00% while the DL-CNN model achieved an accuracy of 98.75%. It is also compared with other existing models like AlexNet, AdaBoostM1 and classifiers such as naïve Bayes, KNN, SVM and random forest classifier models. This model will be useful for provincial hospitals in diagnosing patients, it will also help radiologists in providing more accurate and faster diagnosis by reducing human errors.
Keywords: computer-aided diagnosis; DL-CNN; deep learning convolutional neural network; liver segmentation; image processing.
Remote homology detection using GA and NSGA-II on physicochemical properties
by Mukti Routray, Niranjan Kumar Ray
Abstract: Remote homology detection at amino acid level is a complex problem in the area of computational biology. We have used machine learning algorithms to predict the homology of un-annotated protein sequences which can save time and cost. This work is divided in three phases. Initially the features are extracted from protein sequences using Principal Component Analysis (PCA) to build a chromosome set with representative features of each protein based on physicochemical properties. Second stage involves GA for the construction of a set of chromosomes for classification based on PCA and initialises the classifier to build up an error matrix. Third stage uses NSGA-II, crossover and mutation, and tournament selection for the next set of chromosomes. The output of this experiment is a set of minimum classification error values and minimum number of features used for classification of protein families. This approach gives superior accuracy over the profile-based methods.
Keywords: PCA; principal component analysis; feature selection and classification; genetic algorithm; profile-based methods.
Orthogonal matching pursuit-based feature selection for motor-imagery EEG signal classification
by Rajdeep Chatterjee, Ankita Chatterjee
Abstract: This paper focuses on a framework that uses a small number of features to obtain high-quality classification accuracy of left/right-hand movement motor-imagery EEG signal. Motor-imagery EEG signal has been filtered, and suitable features are extracted using a temporal sliding window-based approach. These extracted features from overlapping and non-overlapping approaches are further compared based on three different types of feature extraction techniques: band power, wavelet energy entropy, and adaptive autoregressive model. The overlapping segments with wavelet energy entropy provide the best classification accuracy over other alternatives. The obtained classification accuracy is 91.43%, the highest ever reported accuracy for BCI Competition II data set III. Subsequently, Orthogonal Matching Pursuit (OMP) technique is used to select the subset of most discriminating features from the entire feature-set. It reduces the computation cost but still retains the quality of the classification results with only 1.43% information loss (that is, 90% classification accuracy), whereas the features-set size reduction is 75% for the same. It is found that the wavelet energy entropy technique performs consistently well in all the variants of our experiments and obtains a mean accuracy difference of 0.95% only.
Keywords: brain-computer interfaces; EEG; ensemble learning; orthogonal matching pursuit; motor-imagery.
The application of plug-and-play ADMM framework and BM3D denoiser for compressed sensing magnetic resonance image reconstruction
by Xiaojun Yuan, Mingfeng Jiang, Lingyan Zhu, Yang Li, Yongming Li, Pin Wang, Tie-Qiang Li
Abstract: Compressed Sensing Magnetic Resonance Imaging (CS-MRI) is an effective technique to reduce MRI data acquisition time. There is currently growing interest in using alternating direction method of multiplier (ADMM) for CS-MRI reconstruction. In this paper, we propose a flexible plug-and-play framework to incorporate the block matching 3D (BM3D) denoising algorithm as prior into the plug-and-play ADMM reconstruction procedure for CS-MRI reconstruction, termed BM3D plug-and-play ADMM (BPA) method. We investigated the performance of the proposed BPA method for the construction of highly under-sampled MRI data of two different sampling masks. Compared with other widely used CS-MRI reconstruction methods, such as PANO, BM3D-IT, BM3D-MRI and BM3D-AMP-MRI, the proposed framework can reconstruct highly under-sampled CS-MRI data with improved gains in peak signal-to-noise ratio and structural similarity index measure.
Keywords: magnetic resonance image reconstruction; plug-and-play ADMM; denoising algorithm; compressed sensing.
Special Issue on: Intelligent Healthcare Systems for Sustainable Development
Dermoscopic image segmentation method based on convolutional neural networks
by Dang N. H. Thanh, Le Thi Thanh, Ugur Erkan, Aditya Khamparia, V. B. Surya Prasath
Abstract: In this paper, we present an efficient dermoscopic image segmentation method based on the linearisation of gamma-correction and convolutional neural networks. Linearisation of gamma-correction is helpful to enhance low-intensity regions of skin lesion areas. Therefore, postprocessing tasks can work more effectively. The proposed convolutional neural network architecture for the segmentation method is based on the VGG-19 network. The acquired training results are convenient to apply the semantic segmentation method. Experimental results are conducted on the public ISIC-2017 dataset. To assess the quality of obtained segmentations, we make use of standard error metrics, such as the Jaccard and Dice, which are based on the overlap with ground truth, along with other measures such as the accuracy, sensitivity, and specificity. Moreover, we provide a comparison of our segmentation results with other similar methods. From experimental results, we infer that our method obtains excellent results in all the metrics and obtains competitive performance over other current and state of the art models for dermoscopic image segmentation.
Keywords: dermoscopic images; deep CNNs; machine learning; skin lesions; image segmentation; skin cancer.
Prediction of diabetic patients using various machine learning techniques
by Shalli RANI, Manpreet Kaur, Deepali Gupta, Amit Kumar Manocha
Abstract: The growth of technology and digitisation of several areas has made the world more successful in reaching solutions to remote problems. Large amounts of health records are also available in digital storage. Machine learning plays an important role for uncovering the health issues from the digital records or for diagnosis of various diseases. In this paper, we present the introduction to recommender system (RS) with respect to diabetic patients after the rigorous review of existing literature. An experiment analysis is performed in Python with the help of machine learning classifiers, such as logistic regression, averaged perception, Bayes point, boosted decision tree, neural network, decision forest, two class support vector machine and locally deep support vector machine on Pima Indian Diabetes Database. We conducted an experiment on 23K diabetic patients dataset. The results from all the classifiers reveal that the logistic regression performs best, with an accuracy of 78% and predicting the accurate results with a specificity of 92%.
Keywords: collaborative filtering; diabetic patients; diabetes mellitus; machine learning.
Multisensor fusion approach: a case study on human physiological factor-based emotion recognition and classification
by A. Reyana, P. Vijayalakshmi, Sandeep Kautish
Abstract: In people's daily life, human emotion plays an essential role, and the mental state accompanied by physiological changes. Experts have always seen that monitoring the perception of emotional changes at an early stage is a matter of concern. Within the next few years, emotion recognition and classification is destined to become an important component in human-machine interaction. Today's medical field makes much use of physiological signals for detection of heart sounds and identifying heart diseases. Thus the parameters temperature and heartbeat can identify the major health risks. This paper takes a new look at the development of an emotion recognition system using physiological signals. In this context, the signals are obtained from the body sensors such as muscle pressure sensor, heartbeat sensor, accelerometer, and capacitive sensor. The emotions observed are happy (excited), sad, angry, and neutral (relaxed). The results of the proposed system shows an accuracy percentage for the emotional states as follows: happy 80%, sad 70%, angry 90%, and neutral 100%.
Keywords: emotion; recognition; multisensor fusion; body sensors; mental state.
LabVIEW based cardiac risk assessment of fetal ECG signal extracted from maternal abdominal signal
by Prabhjot Kaur, Lillie Dewan
Abstract: In recent years, the inclination toward the automated analysis of the fetal ECG signal has become a trend. Mathematical computational processing of abdominal fetal ECG has proved to be beneficial in the crucial diagnosis of complex cardiac diseases. To arrive at the diagnosis, a cardiologist needs to observe the variations critically in the duration and amplitude of different waves and segments of the ECG. In the case of a fetus, a preliminary diagnosis of these deviations helps to have a valid and appropriate intervention, which may otherwise result in permanent damage to the brain and nervous system. For this reason, the fetal cardiac signal has been efficiently extracted from a composite abdominal signal in this paper. The signal extraction has been accomplished in the LabVIEW environment using the Independent Component Analysis (ICA) approach, implemented after the application of hybrid filters, employed for removing noise and artifacts within the signal taken from PhysioNet Database. By proper selection of the cut-off frequency of filters, the denoised signal is approximately 99% accurate. Statistical features, such as the signal-to-noise ratio, standard deviation, error, and accuracy, have been computed as well as morphological features including heart rate, time and amplitude of QRS complex with a duration of the PR interval, RR interval, and QT interval. Results obtained demonstrate that the implementation of ICA for fetal ECG signal extraction helps to determine fetal heart rate accurately with low computational complexity. The performance of the proposed algorithm has also been explored in the case of twin pregnancy. The estimated heart rate is comparable to the actual heart rate, which validates the algorithm's accuracy. The results also indicate the feasibility of real-time application of data acquisition and analysis.
Keywords: electrocardiogram; independent component analysis; sinus rhythm; tachycardia; bradycardia; denoising filters; signal-to-noise ratio; standard deviation; accuracy; LabVIEW.
Impact of feature extraction techniques on cardiac arrhythmia classification: experimental approach
by Manisha Jangra, Sanjeev Kumar Dhull, Krishna Kant Singh
Abstract: This paper provides comparative analysis of state-of-the-art feature extraction techniques in the context of ECG arrhythmia classification. In addition, the authors examine a linear heuristic function LW-index as an indirect measure for separability of feature sets. Seven feature sets are extracted using state-of-the-art feature extraction techniques. These include temporal features, morphological features, EMD-based features, wavelet transform based features, DCT features, Hjorth parameters, and convolutional features. The feature sets performance is evaluated using SVM classifier. The experimental setup is designed to classify ECG signals into four types of arrhythmic beat, which are normal (N), ventricular ectopic beat (VEB), supraventricular ectopic beat (SVEB) and fusion beat (F). A PSO-based feature selection method is used for dimensionality reduction using the LW-index as cost function. The results validate the hypothesis that convolutional features have better discrimination capability as compared with other state-of-the-art features. This paper can resolve the hassles for new researchers related to performance efficacy of individual feature extraction techniques. The work offers an inexpensive methodology and measure to indirectly evaluate and compare the performance of feature sets.
Keywords: ECG; feature extraction; validity index; feature selection; CNN; PSO; DWT; DCT; Hjorth parameters; EMD; temporal features; MIT-BIH database; SVM.
IoT-based automatic intravenous fluid monitoring system for smart medical environment
by Harsha Chauhan, Vishal Verma, Deepali Gupta, Sheifali Gupta
Abstract: Over the last few years, hospitals and other healthcare centres are adopting advances in many sophisticated technologies in order to assure the fast recovery of patients. In almost all hospitals, a caretaker/nurse is responsible for the monitoring of intravenous fluid levels. Usually most of the caretakers forget to change the bottle at the correct time owing to their busy schedule, as a result of which the patient may face problems of reverse flow of blood towards the bottle. To overcome this critical issue, this paper proposes an IoT-based automatic intravenous fluid monitoring system. The proposed device consists of Arduino UNO (i.e. ATMega328 microcontroller), liquid crystal display, solenoid actuator, force sensitive resistor 0.5, ESP8266, a buzzer and LED lights. The authors have used FSR (Force Sensitive Resistor) sensor to monitor the weight of bottle. With the installation of the proposed device, the constant need of monitoring will be reduced by the staff, especially during night hours, thus decreasing the chance to harm the patient and increase the accuracy of healthcare in hospitals. Also, this system will avoid the fatal risk of air embolisms entering the patients bloodstream, which leads to immediate death. To analyse the performance of the proposed system, the authors have done a sample test, by taking time as a parameter to analyse how much time the intravenous fluid bottle is taking to get empty. The results have shown a promising future aspect of the proposed device in order to enhance the healthcare services.
Keywords: drip monitoring system; IoT; healthcare; intravenous fluid; wearable electronics; ESP8266; FSR sensor.
Artificial intelligence based algorithm to track the probable COVID-19 cases using contact history of virus-infected persons
by Javed Shaikh, R.S. Singh, Demissie Jobir Gelmecha, Tadesse Hailu Ayane
Abstract: Currently, the world is facing major challenges in tackling COVID-19. It has affected many countries of the world in terms of human lives, economy and so many other aspects. Many organisations and scientists are working to find ways in which the spread of the COVID-19 can be minimised. One technology that can be effective in tackling this virus is Artificial Intelligence (AI), which can help in many ways. The foremost requirement of this situation is to find the cases of infection as early as possible so that it will not spread rapidly. In this paper, an AI-based algorithm is proposed for the tracking of probable COVID-19 cases. The algorithm uses the mobile numbers of coronavirus-infected persons as data for forecasting. This technique will find the probable infected cases and help in controlling the rapid spread of the virus. This method will provide information regarding an infected person who had contact with other persons by using a forecasting method. As this is an automated tracking system it will help in finding the probable virus-infected cases in a very short time.
Keywords: COVID-19; artificial intelligence; machine learning; forecasting methods.
Prevention of utopsy by establishing a cause-effect relationship between pulmonary embolism and heart failure using machine learning
by Naira Firdous, Sushil Bhardwaj, Amjad Hussain Bhat
Abstract: This paper presents a cause-effect relationship between heart failure and pulmonary embolism, using machine learning. The proposed method is divided into two parts. The first part includes the establishment of connectivity between the two medical fields, which is done by finding out the relationship between the pulse pressure and the stroke volume. The second phase includes the implementation of machine learning on the above-formed connectivity. A univariate technique of feature selection is performed initially in order to get the most relevant attributes. The overfitting problem has been addressed by formulating an ensemble model using hard and soft voting classifiers. Also, the efficiency has been checked by increasing the number of hidden layers of a neural network.
Keywords: pulmonary embolism; stroke volume; pulse pressure; systolic; diastolic; overfitting; ensemble classifiers; neural network.
Tool-based persona for designing user interfaces in healthcare
by Hanaa Alzahrani, Reem Alnanih
Abstract: Technology devices such as smartphones, tablets, and computers have become an intrinsic part of modern life, as this form of technology has entered all businesses and fields such as healthcare. Health sites (HSs) impact healthcare delivery by using technology to improve healthcare outcomes, reduce costs and errors, and increase patient and information safety. Among the available website builders, none has been developed for healthcare sites or designed based on healthcare persona. This is a challenge when designing a specific HS for a particular target group of users such as doctors. System complexity, and the difficulty for doctors to deal with those systems, made it necessary to consider persona that help to understand the mental language of the target users, making the whole systemic experience quite human. The purpose of this paper is to create a new health site design (HSD) tool for designing a User Interface (UI) based User Experience (UX). The tool is designed based on doctors behaviour, personae and real-life scenarios. The applicability of this tool is explored as well as its usability, especially for those with no background in web design. The tool was tested by participants from designing perspectives randomly divided into two groups: control group, who were asked to follow all the instructions in terms of watching and attending the tutorial session and then perform the tasks; and study group, who were asked to perform the tasks directly. The study results show that there is no significant difference between participants in the two groups for effectiveness and efficiency. However, for the cognitive load, the study group was better than the control group. All of the participants were able to complete all the tasks successfully with a minimum amount of time, clicks, and errors. In addition, user satisfaction yielded a score of 84.6 on the System Usability Scale (SUS), mapping it in the A Grade.
Keywords: health systems design tool; website builders; user experience; persona; experimental design; usability evaluation; system usability scale.
RC-DBSCAN: redundancy controlled DBSCAN algorithms for densely deployed wireless sensor network to prolong the network lifespan
by Tripti Sharma, Amar Mohapatra, Geetam Singh Tomar
Abstract: In a wireless sensor network, the nodes are spatially distributed and spread over application-specific experimental fields. The primary role of these nodes is to gather the information for various intended fields such as sound, temperature, vibration, etc. In this proposed algorithm efforts have been made to prolong the network lifespan by decreasing the nodes' energy consumption by considering the critical issues of dense deployment. Every node will limit its chance of participation in any cluster based on the local sensor density. The network area is divided into high- and low-density regions using the DBSCAN algorithm. The nodes in low-density areas are considered critical because there is very little probability for sensing and broadcasting the redundant data by these nodes. The division of high- and low-density regions by applying DBSCAN helps in sleep management. Sleep management helps in energy optimisation in dense areas and thus prolongs network lifetime with the improved stable region. It has been observed through computer simulation that RC-DBSCAN is more energy-efficient than IC-ACO and LEACH in densely deployed network areas in terms of total data packets received by the base station, prolonged network lifespan and improved stability period.
Keywords: DBSCAN; WSN; fuzzy; sleep management.
Coronary artery disease diagnosis using extra tree support vector machine: ET-SVMRBF
by Pooja Rani, Rajneesh Kumar, Anurag Jain
Abstract: Coronary artery disease (CAD) is a type of cardiovascular disease that can lead to cardiac arrest if not diagnosed timely. Angiography is a standard method adopted to diagnose CAD. This method is an invasive method having certain side-effects. So there is a need for non-invasive methods to diagnose CAD using clinical data. In this paper, the authors propose a methodology ET-SVMRBF (Extra Tree SVM-RBF) to diagnose CAD using clinical data. The Z-Alizadeh Sani CAD dataset available on UCI (University of California, Irvine) has been used for validating this methodology. The class imbalance problem in this dataset has been resolved using SMOTE (Synthetic Minority OverSampling Technique). Relevant features are selected using the extra tree feature selection method. The performances of different classifiers XGBoost (Extreme Gradient Boosting), KNN (K-Nearest Neighbour), SVM-Linear (Support Vector Machine-Linear), and SVM-RBF (Support Vector Machine-Radial Basis Function) on the dataset have been evaluated. GridSearch optimisation method was used for hyperparameter optimisation. Accuracy of 95.16% was achieved by ET-SVMRBF, which is higher than recent existing work in the literature.
Keywords: coronary artery disease; cardiovascular disease; extra tree; support vector machine; XGBoost; K-nearest neighbour.
Prediction of cardiac disease using online extreme learning machine
by Sulekha Saxena, Vijay Kumar Gupta, P.N. Hrisheekesha, R.S. Singh
Abstract: This paper presents an automated machine learning (ML) algorithm to detect the coronary disease-like congestive heart failure (CHF) and coronary artery disease (CAD). The proposed automated ML has been employed as a combination of nonlinear features extraction methods, online sequential machine (OS-ELM) and linear discriminate analysis (LDA) as well as generalised discriminate analysis (GDA) as feature reduction algorithms. The dimension reduction of nonlinear features was done by LDA and GDA with Gaussian or radial basis function kernel (RBF), and OSELM binary classifier with an activation function, such as Sigmoid, Hardlim, or RBF, has been used to detect CHF and CAD subjects. For training and validation of ML, twelve nonlinear features were extracted from heart rate variability (HRV) signals. The HRV standard databases have obtained from normal young and elderly CHF and CAD subjects. The numerical experiments were carried out on the sets as CAD-CHF, young-elderly-CAD and young-elderly-CHF subjects. The numerical simulation results clearly have shown when GDA with Gaussian or RBF kernel function is combined with OS-ELM having Sigmoid, Hardlim and RBF activation function, the proposed scheme achieved better detection performance compared with OSELM. To test the robustness of proposed method the classification performances including accuracy, positive prediction value, sensitivity and specificity were calculated on a 100 trial, and it achieved average performance accuracy of 99.77% for young-elderly-CAD and 100% overall performance for CAD-CHF and young-elderly-CHF subjects.
Keywords: Lempel-Ziv; Poincare plot; OSELM; sample entropy; dimension reduction method; detrended fluctuation analysis.
Digitisation of paper-ECG using column-median approach
by Priyanka Gautam, Ramesh Kumar Sunkaria, Lakhan Dev Sharma
Abstract: Usually, ECG (electrocardiogram) signals are recorded on standard grid paper to determine the potential of cardiac disorders in hospitals. In the current technological era, existing records of paper-ECG are needed to be converted into digital forms as it is the most effective technique to analyse, process, store and communicate attributes of ECG (features/quality, etc.) for clinical uses. The present work introduces a novel technique for the digitisation of paper-ECG (column-median approach). This paper uses correlation and heart rate as parameters to validate the proposed methodology. To observe the precision of the proposed algorithm, the accuracy of the heart rate is also calculated. The overall correlation and percentage error carried out in 50 different signals are 0.86 and 0.79%, respectively. The overall accuracy obtained for 50 different ECG signals is 99.21%, which shows that the methodology works effectively.
Keywords: paper-ECG; column-median approach; biomedical image processing.
Special Issue on: The Significance of Machine Learning for COVID-19
Analysis of some topological nodes using the adaptive control based on 9-D, hypothesis theoretical to COVID-19
by Abdulsattar Abdullah Hamad, M.Lellis Thivagar, K. Martin Sagayam
Abstract: This work is an extension based on a new model previously proposed where the Hamiltonian, synchronisation, Lyapunov expansion, equilibrium, and stability of the proposed model for the same authors were analysed. In this paper we present a broader analysis to develop receiving network nodes faster. The analysis and study have demonstrated how to determine the basic structure and content of the Sym in theory, an attempt to identify objects that have a fundamental engineering role for the model after confirming the performance and results, we can suggest it to determine the spread of coronavirus.
Keywords: lu; Hamiltonian; synchronisation; Lyapunov expansion; equilibrium; topological nodes.
An ensemble approach to forecast COVID-19 incidences using linear and nonlinear statistical models
by Asmita Mahajan, Nonita Sharma, Firas Husham AlMukhtar, Monika Mangla, Krishna Pal Sharma, Rajneesh Rani
Abstract: Coronavirus 2019, also known as COVID-19, is currently a global epidemic. This pandemic has infected more than 100 countries all over the globe and is continuously spreading and endangering the human species. Researchers are perpetually trying to discover a permanent antidote for the virus, but presently no particular medication is available. As a result, health sectors worldwide are experiencing an unexpected rise in cases each day. Hence, it becomes necessary to predict the spread of the disease so as to enable public health sectors to improve their control capabilities in order to mitigate the spread of the infections. This manuscript proposes a stacked ensemble model for accurately forecasting the future occurrences of COVID-19. The proposed ensemble model uses Exponential Smoothing (ETS), Autoregressive Integrated Moving Average (ARIMA), and Neural Network Autoregression (NNAR) as the base models. Each base model is trained individually on the disease dataset, whose regress values are then used to train the Multilayer perceptron (MLP) model. The stacked model gives better predictions compared with all the other four forecasting models. It is validated that the proposed model outperforms the base models. This validation is established through error metrics such as Root Mean Square Error (RMSE) and Mean Absolute Error (MAE). The results conclude that the ensemble model is highly robust and reliable in forecasting future COVID incidences in comparison to other statistical time series models.
Keywords: COVID-19; pandemic; forecasting; ensemble approach; stacking; autoregressive integrated moving average; exponential smoothing; neural network; multilayer perceptron.
Simple program for computing objective optical properties of magnetic lenses
by R.Y. J. Al-Salih, Abdullah E. M. AlAbdulla, Ezaldeen Mahmood Abdalla Alkattan
Abstract: This paper describes the basic features of a new program denoted MELOP (Magnetic Electron Lens Optical Properties), primarily intended for providing a free simple way to calculate the objective focal properties of rotationally-symmetric electron magnetic lenses in the presence of an axial magnetic field distribution. The calculation is done by solving the paraxial ray equation using the fourth-order Runge-Kutta formula. For a specific beam voltage, the program computes the excitation parameter, the object or image plane, the objective principal plane, the objective focal length, the objective magnification, the spherical aberration coefficient, the chromatic aberration coefficient, and the magnetic flux density at the object or image plane. These parameters are solved for zero, low, high or infinite magnification condition. The program can handle instantaneously plotting the variations of the calculated parameters relatively. In addition to that, the data can be transferred to xlsx or txt file format.
Keywords: electron lenses design; electron objective focal properties; fourth-order Runge-Kutta formula.
The impact of oil exports on consumer imports in the Iraqi economy during the COVID-19 period: a theoretical study
by Mustafa Kamil Rasheed, Ali Mahdi Abbas Al-Bairmani, Abir Mohammed Jasim Al-Hussaini
Abstract: Exports and imports of foreign trade are widely considered to be the most important contribution to the economic development of society. Especially, the potential and competitiveness of exports are realised, result from that an import capacity that supports growth and balance in all economic sectors. Particularly, the foreign exchange revenues come up with increasing exports, which tend to finance investment projects as well as encourage the importation of developed means of production that contribute to increase productivity and achieve economic efficiency, but this is rarely achieved in developing countries including Iraq. In spite of the high amount of oil exports, there are a large proportion of revenues from these exports that go towards import consumer goods, hence do not create a stimulating environment for production and investment. On the contrary, they stimulate the investment multiplier in the exporting partner countries, which stimulates their investment activity. The hypothesis of this study refers to how the direct relationship between oil exports and consumer imports disrupts the economy and output and weakens its performance. The most important finding of the study is that oil exports in Iraq directly link with consumer imports, which leads to the loss of the Iraqi economy's financial resource and stimulating economic activity. The study recommends the need to adopt economic diversification to overcome the unilateral Iraqi economy, as well as the optimal use of financial resources to support the national economy.
Keywords: oil exports; consumer imports; total exports; production activities; COVID-19.
Evaluation of the impact parameters of nano Al2O3 dielectric in wire cut-electrical release machining in the COVID-19 environment
by Farook Nehad Abed, Azwan Bin Sapit, Saad Kariem Shather
Abstract: This paper focuses on wire electric discharge machine in the COVID-19 environment. It can be considered as an attempt to develop models of response variables. Using a different liquid, one of which is nanoparticle (Al2O3), in the ratio of (2 mg) and the function of comparison in both cases is the rate of material removal, in the wire electric discharge machine process using the response surface methodology. The pilot plan is based on the concept of the Box-Behnken, and the study conveys the six main parameters. To evaluate the value of the advanced model, ANOVA was applied; the test results support the validity and suitability of the advanced RSM model. Optimum settings for the parameters are improved work safety in the COVID-19 environment.
Keywords: wire electric discharge machine; titanium; MRR; RSM; COVID-19.
Analysis of convolutional recurrent neural network classifier for COVID-19 symptoms over computerised tomography images
by Srihari Kannan, N. Yuvaraj, Barzan Abdulazeez Idrees, P. Arulprakash, Vijayakumar Ranganathan, Udayakumar E., P. Dhinakar
Abstract: In this paper, a Convolutional Recurrent Neural Network (CRNN) model is designed to classify the patients with COVID-19 infections. The CRNN model is designed to identify the Computerised Tomography (CT) images. The processing of CRNN is modelled with input image processing and feature extraction using CNN and prediction by RNN model that quickens the entire process. The simulation is carried with a set of 226 CT images by varying the training-testing accuracy on a 10-fold cross-validation. The accuracy in estimating the image samples is increased with increased training data. The results of the simulation show that the proposed method has higher accuracy and reduced MSE with higher training data than other methods.
Keywords: image classification; COVID-19; medical imaging; convolutional recurrent neural network; 10-fold cross-validation.
An empirical validation of learning from home: a case study of COVID-19 catalysed online distance learning in India and Morocco
by Gabriel A. Ogunmola, Wegayehu Enbeyele, Wissale Mahdaoui
Abstract: The world as we know it has changed over a short period of time, with the rise and spread of the deadly novel coronavirus known as COVID19 and will never be the same again. This study explores the devastating effects of the novel virus pandemic and the resulting lockdown, thus the need to transform the offline classroom into an online classroom. It explores and describes the numerous online teaching platforms, study materials, techniques, and technologies being used to ensure that educating the students does not stop. Furthermore, it identifies the platforms and technologies that can be used to conduct online examinations in a safe environment devoid of cheating. Additionally, it explores the challenges facing the deployment of online teaching methods. On the basis of literature review, a framework is proposed to deliver superior online classroom experience for the students, so that online classroom is as effective as or even better than offline classrooms. The identified variables were empirically tested with the aid of a structured questionnaire: there were 340 respondents who were purposefully sampled. The result indicates that students prefer online teaching when such sessions are enhanced with multimedia presentations. The study recommends that instructors need to train in the use of technology enhanced learning if learning from home is going to be effective.
Keywords: COVID–19; online classroom; Zoom; lockdown; MOOC; iCloud; proportional odds model.
An empirical study on social contact tracing of COVID-19 from a classification erspective
by Mohammed Gouse Galety, Elham Tahsin Yasin, Abdellah Behri Awol, Lubab Talib
Abstract: The staggering emergency of COVID-19 is a pandemic and irresistible without the antibody and cure. This uprising issue needs the preventive controls through creation of awareness and implementation of the contact tracing process. The procedure of contact tracing is to determine the infected individual or the men and women who have had contact with infected people to be indexed and dealt with carefully. This device has its usage to lessen the infections with the information described at the infectious disease and to reduce the spread of the infection with precautionary measures by means of creating awareness. Awareness creation demands various tools for its installation whereas social media networking is a knowledge set of the current market coverage of maximum sociology of the planet to see, analyse and interpret the present the market knowledge with the support of classifiers and applied math learning ways of Artificial Intelligence (AI). This paper derives the obtainable data with its analysis and widespread contract tracing through the use of social media dataset, similarly as math learning ways of AI are applied to determine the infected COVID-19 and infers the adequate action and preventive measures for the reduction of the expansion of COVID19 infections by the controller's segment of the said method.
Keywords: COVID-19; infection; preventive controls; awareness creation; social media; contact tracing; artificial intelligence.
Analysis of the COVID-19 pandemic and forecasting using machine learning models
by Ekansh Chauhan, Manpreet Sirswal, Deepak Gupta, Ashish Khanna, Aditya Khamparia
Abstract: The coronavirus pandemic is rapid and universal, menacing thousands of lives and all economies. The full analysis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is imperative as a deciding factor for the remedic actions. Machine learning is being used in every sphere to fight the coronavirus, be it understanding the biology of the virus in time, be it diagnosing the patients, or be it drug and vaccine development. It is also critical to predict the pandemic lifetime so to decide on opportune and remedic activities. Being able to accurately forecast the fate of an epidemic is a critical but difficult task. In this paper, based on public data available to the world and India, the estimation of pandemic parameters and the ten days ahead forecast of the coronavirus cases is proposed using Prophet, Polynomial Regression, Auto Arima and Support Vector Machine (SVM). The performances of all the models were motivating. MAE and RMSE of polynomial regression and SVM were convincingly low. Polynomial regression has predicted the highest number of cases for India and the lowest number of cases for the world, which depicts that according to polynomial regression the daily cases are going to spike in India and decline a little in the world. Prophet has forecast the lowest number of cases for India and the highest number of cases in the world, after SVM. The results of Arima are closest to the average of combined results by all of the four models. The only limitation is the lack of enough data, which creates high uncertainty in the forecast. The four factors, i.e. growth factor, growth ratio, growth rate and second derivative for the growth of coronavirus, in the USA and India are also calculated and compared. Several theories revolving around the origin of the coronavirus are also discussed in this paper. Under optimistic predictions, the results show that the pandemic in some countries is going to terminate soon, while in some countries it is going to increase at an alarming rate and the overall rate of growth of the coronavirus cases is decreasing in both the USA and India.
Keywords: COVID-19; machine learning; novel coronavirus; classification; Technology.
A statistical analysis for COVID-19 as a contact tracing approach and social networking communication management
by Abdulsattar A. Hamad, Anasuya Swain, Suneeta Satpathy, Saibal Dutta
Abstract: The COVID19 outbreak caused by severe acute respiratory syndrome coronavirus 2 (SARSCoV2) has been declared as a global pandemic. The first case of coronavirus was detected in Wuhan city of China and later declared as pandemic by the World Health Organization. As of the first week of August 2020, more than 20 million cases of COVID19 have been reported globally, resulting in more than 700,000 deaths and around 12 million people have recovered. The medium of spread of viral infection is the droplets produced from nose and mouth by coughing, sneezing and talking or by small droplets which hang in the air. COVID-19 disease as yet has no vaccine or medication. Preventive measures for this infectious disease may include creating awareness and implementation of the contact tracing process. The process of contract tracing is to determine the infected person or the people who have had contact with infected people, to be listed and treated carefully. The basic aim is to reduce the infections with the detailed description of COVID-19 and minimise the spreading of the infections by creating awareness. Awareness creation demands the adoption of different tools, among which social media networking is considered as a fruitful medium to achieve the same. Further different classifiers and statistical learning methods are also used to analyse and interpret the social media networking data. This research study has derived the available information with the employment of statistical learning methods of artificial intelligence and successful contract tracing through the use of social media datasets to determine the infected COVID-19 data. In addition, this research also infers the adequate course of action and preventive measures for the reduction of the growth of COVID19 infections with the help of the controllers segment of the said method. The present work has also adopted the Natural Language Processing (NLP) method as an aid to process the social network data and find the solution to the inquiries. In addition, the work has also validated the relationship between the social networking and employment of artificial intelligence techniques as a contact tracing and awareness program with the help of statistical tools like regression, coefficient correlation and Anova. The main objective of the study is to reduce the pandemic infections by spreading awareness and generating detailed descriptive reports about COVID-19 with the usage of social media networking as well as artificial intelligence statistical learning methods.
Keywords: COVID-19; infection; preventive controls; awareness creation; social media; contact tracing; artificial intelligence.
The degree of applying electronic learning in the Gifted School of Nineveh in Iraq, and what management provided to the students and its relationship to qualitative education during the COVID-19 pandemic.
by Ahmed S. Al-Obeidi, Nawar A. Sultan, Anas R. Obaid, Abdulsattar A. Hamad
Abstract: This paper discusses the most important pillars of e-learning and the distance learning process in a Gifted School in Nineveh. Through this study, we were able to identify the methods of conducting distance education under the information technology system and on the work and learning environment used in e-learning. Increasing the efficiency of the educational institution through distance e-learning just as the basics of building an e-learning system in various educational institutions. The types of program and the best-known of which are in the application of e-learning in educational institutions in general and in the Gifted School in particular are also discussed. A comparison is made between the two in terms of method and accuracy of using these programs.
Keywords: electronic learning; Gifted School of Nineveh; COVID-19; distance education; hypothetical education.
Design and analysis on the molecular level of a biomedical event trigger extraction using recurrent neural network based particle swarm optimisation for COVID-19 research
by R.N. Devendra Kumar, Arvind Chakrapani, Srihari Kannan
Abstract: In this paper, rich extracted feature sets are fed to the deep learning classifier that estimates the optimal extraction of lung molecule triggered events for COVID-19 infections. The feature extraction is carried out using a Recurrent Neural Network (RNN) that effectively extracts the features from the rich datasets. Secondly, a particle swarm optimisation (PSO) algorithm is used to classify the extracted features of COVID-19 infections. The rule set for the feature extractor is supplied by the fuzzy logic rule set. The simulation shows that the RNN-PSO, which is the combination of two algorithms, offers improved performance over other machine learning classifiers.
Keywords: event triggers; COVID-19; lung molecules; feature extraction; classification; particle swarm optimisation; recurrent neural network.
Multivariate economic analysis of the government policies and COVID-19 on the financial sector
by Monika Mangla, Nonita Sharma, Sourabh Yadav, Vaishali Mehta, Deepti Kakkar, Prabakar Kandukuri
Abstract: The whole world is experiencing a sudden pandemic outbreak of COVID-19. In the absence of any specific treatment or vaccine, social distancing has proved to be an effective strategy in containing the outbreak. However, this has led to disruption in trade, travel, and commerce by halting manufacturing industries, by the closing of corporate offices, and all other sundry activities. The alarming pace of the virus spread and the increased uncertainty is quite concerning to the leading financial stakeholders. This has led to the customers, investors, and foreign trading partners fleeing away from new investments. Global markets plummeted, leading to erosion of more than the US $6 trillion within just one week from 24 to 28 February 2020. During the same week, the S&P 500 index alone experienced a loss of more than $5 trillion in the USA, while other top 10 companies in the S&P 500 suffered a combined loss of more than $1.4 trillion. This manuscript performs multivariate analysis of the financial markets during the COVID-19 period and thus correlates its impact on the worldwide economy. An empirical evaluation of the effect of containment policies on financial activity, stock market indices, purchasing manager index, and commodity prices is also carried out. The obtained results reveal that the number of lockdown days, fiscal stimulus, and overseas travel ban significantly influence the level of economic activity.
Keywords: coronavirus; COVID-19; financial sector; forecasting; multivariate analysis; NIFTY indices; pandemic; regression model; stringency index.
COVID-19 suspected person detection and identification using thermal imaging based closed circuit television camera and tracking using drone in internet of things
by Pawan Singh Mehra, Yogita Bisht Mehra, Arvind Dagur, Anshu Kumar Dwivedi, M.N. Doja, Aatif Jamshed
Abstract: COVID-19 has emerged as a world-wide health concern where human to human transmission is described with an incubation period of 2-10 days. It is contagion by droplet and contaminated surfaces like hands. The sole way to detect a suspected person being infected with COVID-19 without COVID-19 testing kit is through thermal scanner. Since the disease is spreading at a vast rate, not only it is very hard to check or scan every individual manually but also there are chances of transmission of COVID-19 to the unsuspecting person. In this paper, we propound a system where the suspected person can be easily detected and identified for COVID-19 by using thermal imaging based closed circuit television (CCTV), which will automatically scan the people in the vicinity and capture a video/image of the suspected person. The system will raise an alarm in the vicinity so that people in that area can distance themselves from each other. The recorded video/image will be forwarded to base station and information about the suspected person will be fetched from the server. Meanwhile, drones will be used for tracking the suspected person until the nodal medical team diagnose the suspected person for confirmation. The proposed system can contribute significantly for curbing the rate of infected COVID-19 persons and prevent further spread of this pandemic disease.
Keywords: coronavirus; COVID-19; face recognition; drone; internet of things; automation; deep learning.
Machine learning based classification: an analysis based on COVID-19 transmission electron microscopy images
by Kalyan Kumar Jena, Sourav Kumar Bhoi, Soumya Ranjan Nayak, Chinmaya Ranjan Pattanaik
Abstract: A virus is a type of microorganism which has an adverse effect on human society. Viruses replicate themselves within the human cells rapidly. Currently, the effects of very dangerous infectious viruses are a major issue throughout the globe. Coronavirus (CV) is considered as one of the dangerous infectious viruses for the entire world. So, it is very important to detect and classify this type of virus at the initial stage so that preventive measures can be taken as early as possible. In this work, a machine learning (ML) based approach is used for the type classification of CV such as alpha CV (ACV), beta CV (BCV) and gamma CV (GCV). The ML-based approach mainly focuses on several classification techniques, such as support vector machine (SVM), Random Forest (RF), AdaBoost (AB) and Decision Tree (DT) techniques by processing several CV images (CVIs). The performance of these techniques is analysed using a classification accuracy performance metric. The simulation of this work is carried out using Orange3-3.24.1.
Keywords: COVID-19; machine learning; TEM CVIs; support vector machine; random forest; AdaBoost; decision tree.
Gradient and statistical features based prediction system for COVID-19 using chest X-ray images
by Anurag Jain, Shamik Tiwari, Tanupriya Choudhury, Bhupesh Kumar Dewangan
Abstract: As per data available on the WHO website, the count of COVID-19 patients on 20 June 2020 had surpassed the figure of 8.7 million globally and around 460,000 had lost their lives. The most common diagnostic test for COVID-19 detection is a Polymerase Chain Reaction (PCR) test. In highly populated developing countries such as Brazil, India etc., there has been a severe shortage of PCR test-kits. Furthermore, the PCR-test is very specific and has low sensitivity. This implies that the test can be negative even when the patient is infected. Moreover, it is expensive too. While efforts to intensify the volume and accuracy for PCR testing are in progress, medical practitioners are trying to develop alternative systems using medical imaging in the form of chest radiography or CT scans. In this research work, we have preferred chest X-rays for COVID-19 detection owing to wide availability of chest X-ray infrastructure in all over world. We have designed a decision support system based on statistical features and edge maps of X-ray images to detect COVID-19 virus in a patient. Online available datasets of chest X-ray images have been used to train and test decision tree, K-nearest neighbour, random forest, and multilayer perceptron machine learning classifiers. From the experimental results, it has been found that the multilayer perceptron achieved 94% accuracy, which is highest among the four classifiers.
Keywords: COVID-19; chest X-ray; statistical features; image gradient; random forest; KNN; multilayer perceptron; decision tree.
Indian COVID-19 time series prediction using Facebooks Prophet model
by Mamata Garanayak, Goutam Sahu, Mohammad Gouse Baig, Sujata Chakravarty
Abstract: The entire world has been facing an unprecedented public health crisis due to COVID-19 pandemic for the last year. Meanwhile, more than one million people across the world have already died; many more millions are under treatment. Some countries in Europe have begun to experience the second wave of the pandemic too. This has put the entire health infrastructure of countries under severe strain and has led to downward spiral in the economy. The most worrisome part is the uncertainty as to the spread or arrest of the pandemic. In such a scenario, robust forecasting methods are needed to enable health professionals and governments to make necessary preparation in accordance with the situation. Artificial intelligence and machine learning techniques are useful tools not only for collection of accurate data but also for prediction. Studies show that time series forecasting techniques, such as Facebooks Prophet, have shown promising results. In this paper, time series techniques have been used to forecast the numbers of deaths, recovery and positive cases 60 days ahead. The experimental results demonstrate that machine learning techniques can be beneficial in forecasting the behaviour of the pandemic.
Keywords: machine learning; Prophet; COVID-19; time series; coronavirus; prediction.
Transmission dynamics of COVID-19 outbreak in India and effectiveness of self-quarantine: a phase-wise data-driven analysis
by Sahil Khan, Md. Wasim Khan, Narendra Kumar, Ravins Dohare, Shweta Sankhwar
Abstract: The novel coronavirus disease referred as COVID-19 was declared as a pandemic by the World Health Organization. During this pandemic more than 988,172 lives were lost and 7,506,090 approximately active cases were found across the world by 25 September, 2020. To predict the novel coronavirus transmission dynamics in India, the SQEIHDR mathematical model is proposed. The model is an extension of basic SEIR mathematical model with additional compartments. These additional compartments include self-quarantine (Q), isolation (H) and deceased (D), which help to understand the COVID-19 outbreak in India in a more realistic way and is supposed to suppress the rise of transmission. The SQEIHDR model's simulation comprises ten phases (phases 0-9) with different COVID-19 preparedness and response plans. The simulation results show significant changes in the curve of infected population based on variation in compartment Q, which reveals the efficacy of imposed as well as proposed preparedness and response plans. The results of different conditions of preparedness and response plans highlight the key to reduce the outbreak, i.e. the rate of self-quarantine (Q) which includes general awareness, social distancing and food availability.
Keywords: COVID-19; mathematical modelling; self-quarantine; transmission dynamics; preparedness; response plan.
COVID-19 outbreak in Orissa: MLR and H-SVR based modelling and forecasting
by Satyabrata Dash, Hemraj Saini, Sujata Chakravarty
Abstract: WHO declared COVID-19 to be a pandemic in early March, 2020, and by June it became a severe threat to the human community in almost every country. The present situation throughout the world is very tense and puts everyone at a high risk of infection and this further leads to the high mortality rate. Everyone in the related research community is using technology and trying to identify the time at which the pandemic might stop and make the world healthy again. Therefore, in this study, an attempt has been made to analyse and predict COVID-19 outbreak using Multiple Linear Regression (MLR) and Support Vector Regression (SVR). In this comparative analysis, MLR outperforms SVR. Hence, MLR can be used to predict COVID-19 outbreak in the real life applications.
Keywords: novel coronavirus; COVID-19; linear multiple regression; support vector regression.
Prediction of COVID-19 epidemic curve in India using the supervised learning approach
by Shweta Mongia, N. Jaisankar, Sugandha Sharma, Manoj Kumar, Vasudha Arora, Thompson Stephan, Achyut Shankar, Pragya Gupta, Raghav Kachhawaha
Abstract: The COVID-19 pandemic, a neo zoonotic infectious disease, has caused high mortality worldwide. The need of the hour is to equip the governments with early detection, prevention, and mitigation of such contagious diseases. In this paper, a supervised learning approach of the polynomial regression model is used for the prediction of COVID-19 cases in terms of the number of Confirmed Cases (CC), Death Cases (DC), and Recovered Cases (RC) in India. As per the prediction model, the epidemic curve will reach its peak on 31 May 2020 when the predicted number of CC (148,276) will be almost equal to the sum of the number of DC (35,050) and RC (114,718) i.e. 149,768. This research is based on the data available till 25 April 2020 considering a strong preventive measure of nation-wide lockdown in India since 24 March 2020. Authors have also predicted death rates and recovery rates. As of 25 April 2020, the death rate stands at 3.068% and the predicted death rate for 1 June 2020 is 2.558%. The recovery rate on 25 April 2020 is 21.97% and it is predicted that by 1 June 2020 this rate will increase to 79%. In addition to this, the approach projected a monthly percentage increase in the number of CC from 1 May 2020 to 1 December 2020. This analysis would help and enable the concerned authorities in bringing effective preventive measures into action in the process of decision making.
Keywords: supervised learning; polynomial regression model; COVID-19; prediction; epidemic curve.
Special Issue on: Signal and Information Processing in Sensor and Transducer Systems
Identification of Hammerstein-Wiener nonlinear dynamic models using conjugate gradient based iterative algorithm
by Xiangli Li, Lincheng Zhou
Abstract: This paper mainly studies the identification of a class of nonlinear dynamic models with Hammerstein-Wiener nonlinearity.Firstly, a special form of Hammerstein-Wiener polynomial model is constructed by using the key term decomposition technique to separate the model parameters to be estimated.On this basis, an iterative algorithm based on conjugate gradient (CGI) is proposed, which computes a new conjugate vector along the conjugate direction in each iteration step.Because the search direction of the CGI algorithm is conjugate with respect to the Hessian matrix of the cost function, the CGI algorithm can generally obtain the faster convergence rates than the gradient based iterative algorithm.By conjugating the search direction of the CGI algorithm with the Hessian matrix of the loss function, CGI algorithm has more advantages in convergence rates than the gradient based iterative algorithm. Finally, numerical examples are given to demonstrate the effectiveness of the proposed algorithm.
Keywords: Hammerstein-Wiener model; conjugate gradient; key term decomposition; Hessian matrix; parameter estimation.
Multi-sensor temperature and humidity control system of wine cellar based on cooperative control of intelligent vehicle and UAV
by Yufan Wang
Abstract: Red wine has extremely strict requirements on its fermentation and long-term storage environment. The rapid change of cellar temperature will cause great damage to the taste of red wine. At present, wine cellars at home and abroad usually adopt pure manual management or lay a large number of sensors for monitoring to solve such problems. However, in the face of fire risks caused by specialized laboratories or excessive costs and a large number of aging production lines, it is obvious that the needs of wine cellar managers cannot be met. This paper designs and completes the multi-point data acquisition temperature and humidity adjustment sensor system of the wine cellar under the collaborative control of smart car and UAV. The system consists of four independent parts: the intelligent patrol car terminal, the four-rotor UAV auxiliary terminal, the handheld terminal and the intelligent temperature control device terminal, which cooperate with each other to realize the monitoring and control of the environment under closed conditions. Compared with the traditional temperature and humidity collection method, this system uses the method of UAV and smart car to collect and collect data, which greatly improves the efficiency and accuracy of data collection. The system is equipped with low-power autonomous charging to realize unmanned management. At the same time, the administrator can view and intervene in the real-time changes of the indoor environment through the handheld segment to achieve human-computer interaction. Experimental tests show that this system has strong robustness and adaptability, is accurate, intelligent, efficient, and saves a lot of manpower and material resources.
Keywords: temperature and humidity acquisition sensor system; DHT11; cooperation; human-computer interaction
Simulation study on identification technology of transmission line potential hazards based on corona discharge characteristics
by wei liu
Abstract: The prevention and control of transmission line potential hazard is the guarantee of safe and reliable operation of power grid. At present, the prevention and control of line potential hazard is still based on manual inspection, which has problems of low efficiency and poor reliability. Based on corona discharge theory and experimental simulation, this paper studies the fingerprint characteristics of line discharge signal caused by tree barrier, bird damage and insulator pollution, and puts forward a method of line potential hazard detection and fault identification based on discharge characteristics. The results show that the development of line potential hazard will lead to the discharge process, and the discharge characteristics of different types of potential hazards have obvious differences. The differences are mainly reflected in the main wave width and discharge repetition rate, which can be used to identify the types of potential hazards.
Keywords: Overhead line, potential hazard, Corona discharge; Simulation analysis, Identification.
Development and Application of PD Spatial Location System in Distributing Substation
by fang peng, Hong-yu Zhou, Xiao-ming Zhao
Abstract: Partial discharge is an important cause of insulation deterioration of distribution network equipment. Due to the variety of distribution network equipment, the location of discharge source is always a technical difficulty in engineering. In this paper, through the research of UHF PD spatial location technology, a system for spatial location of discharge source in distribution room is developed. The UHF sensor, acquisition and processing module and analysis and diagnosis module for the location of discharge source in distribution room are designed. The results of laboratory test and actual operation show that the system has the advantages of high detection sensitivity, high location accuracy and high operation reliability. It can be used for effective monitoring and timely warning of PD defects in distribution room, which helps to improve the power supply reliability of distribution network system.
Keywords: Distributing substation, Partial discharge, Spatial location, Sensor, Online monitoring.
Feature Matching for Multi-beam Sonar Image Sequence using KD-Tree and KNN Search
by Jue Gao
Abstract: Feature matching for image sequence generated by multi-beam sonar is a critical step in widespread applications like image mosaic, image registration, motion estimation and object tracking. In many cases, feature matching is accomplished by nearest neighbour arithmetic on extracted features, but the global search adopted brings heavy computational burden. Furthermore, sonar imaging characteristics such as low resolution, low SNR, inhomogeneity, point of view changes and other artifacts sometimes lead to poor sonar image quality. This paper presents an approach to the feature extraction, K-Dimension Tree (KD-Tree) construction, and subsequent matching of the features in multi-beam sonar images. Initially, Scale Invariant Feature Transform (SIFT) method are used to extract features. A KD-Tree based on feature location is then constructed. By K Nearest Neighbour (KNN) search, every SIFT feature is matched with K candidates between a pair of consecutive frames. Finally, the Random Sample Consensus (RANSAC) arithmetic is used to eliminate wrong matches. The performances of the proposed approach are assessed with measured data that exhibited reliable results with limited computational burden for the feature-matching task.
Keywords: feature extraction; feature matching; multi-beam sonar; KD-Tree; KNN.
A study on ultrasonic process tomography for dispersed small particle system visualization
by Zhiheng Meng, Jianfei Gu, Yongxin Chou
Abstract: The present challenge in the ultrasonic process tomography on dispersed small particle system is that it is hard to obtain the accurate algorithm to reconstruction. For more accurate reconstruction, this work proposes an improved GMRES(Generalized Minimal Residual)algorithm based on generalized minimal residual iteration and mean filtering method. To verify the feasibility of the algorithm for dispersed small particle system visualization, a linear acoustic attenuation model is developed to obtain the projection data of ultrasonic array. Then, we compared it with the current mainstream reconstruction algorithms under the conditions of the less effective information by solving the underdetermined equations. It is showed that this method can present a high reconstruction precision in the cases of numerical simulations, and reasonably reflect the cross section of dispersed small particle distribution. In the numerical simulations, the imaging accuracy of improved GMRES algorithm can reach about 90%.
Keywords: ultrasonic method; dispersed particle; particulate two-phase flow; back projection; iterative algorithm.
Distributed fusion algorithm based on maximum internal ellipsoid mechanism
by Jinliang Cong
Abstract: In this paper, a Bar-Shalom Campo based algorithm is presented to solve the approximate maximum ellipsoid in the cross region of covariance ellipsoid. An objective function that can be solved by linear matrix inequality is designed based on the rotation transformation of matrix. Compared with the classical covariance intersection fusion algorithm, it is less conservative. Moreover, the unknown cross-covariance is approximated as a linear matrix inequality constraint with Pearson correlation coefficient which is bounded. With the inequality constraint, the accuracy of fusion results can be improved. Finally, two simulation examples are given to verify the effectiveness of the proposed algorithm.
Keywords: distributed sensor network, information fusion, maximum ellipsoid, cross-covariance constraint.
Local Track to Detect for Video Object Detection
by Biao Zeng, Shan Zhong, Lifan Zhou, Zhaohui Wang, Shengrong Gong
Abstract: The existing methods for video object detection are generally achieved from searching the objects through the entire image. However, they always suffer from large computation consumption as a result of dozens of similar images are required to be operated. To relieve this problem, we propose a Local Track to Detect (LTD) framework to detect video objects by predicting the movements of objects in local areas. LTD can automatically determine key frames and non-key frames, the objects in key frames can be detected by the single frame detector, and the objects in non-key frames can be efficiently detected by the movement prediction module. LTD also has a siamese module to predict whether objects between the key frame and the non-key frame are the same object to ensure the accuracy of the movement prediction module. Compared with other previous work, our method is more efficient and achieves state-of-the-art performance.
Keywords: Video Object Detection; Local Detection; Detect and Track; Movement Prediction; Efficient Detection; CNN.