International Journal of Sensor Networks (13 papers in press)

Regular Issues

• A Hazardous Chemical Accident Prevention Method Based on Event Logic Graph  
  by Guanlin Chen, Zhenhua Kong, Wenyong Weng, Qi Lu, Wenfang Zhou 
  Abstract: In recent years, hazardous chemical accidents have occurred frequently at home and abroad, causing huge damage to the personal and property safety of the general public. This paper proposes a hazardous chemical accident prevention method based on an event logic graph associated with a sensor system to predict potential hazardous chemical accidents by extracting events from the accessible material safety data sheet data of hazardous chemicals, extracting event relationships for the extracted events, and inserting the extracted events into the graph database Neo4j. Finally, an early warning system is implemented based on the reasoning of the event map. This system can prevent disasters before they occur. In addition, after a catastrophe has occurred, this system can also analyse the causal chain of the accident, which is of profound significance whether it provides guidance for emergency rescue or disaster relief or inspects and upgrades the warehouse system afterward.
  Keywords: hazardous chemicals; event logic graph; early warning system; sensor.
  DOI: 10.1504/IJSNET.2024.10063256
   
  
• Enhancing Railway Transportation Safety with Proactive Maintenance Strategies Incorporating Machine Learning  
  by Q.U.N. Wei, Ning Zhao 
  Abstract: Recognising the significance of railway infrastructure, effective maintenance is vital to prevent breakdowns, accidents, and ensure smooth operations. This research aims to develop a novel machine learning-based railway predictive maintenance (MLT-RPM) system to address issues of downtime and resource allocation. The system employs sensors to record data on temperature, vibration, and wear, enabling early diagnosis and prevention of locomotive engine failures. This framework enhances railway infrastructure's security and reliability, minimising downtime, costs, and accidents. The study also demonstrates that MLT-RPM reduces energy consumption and environmental impact, promoting safety, dependability, cost savings, operational efficiency, and environmental sustainability.
  Keywords: railway; machine learning; predictive maintenance system; locomotive engine; sensor; resource utilisation; safety; reliability.
  DOI: 10.1504/IJSNET.2024.10063788
   
  
• BRAYOLOv7: An Improved Model Based on Attention Mechanism and Raspberry Pi Implementation for Online Education  
  by Jiayi Wu, Yingqian Zhang, Lei Fu, Yunrong Luo, Hui Xie, Rongru Hua 
  Abstract: Traditional machine learning in the education industry is facing difficulties in accurately identifying students' emotions, impacting the personalised delivery of online education. To address this, we propose the development of an enhanced YOLOv7 model called BRAYOLOv7, which utilises the bi-level routing attention mechanism. Our approach includes adjusting the non-maximal suppression parameter to reduce accidental deletion and false detection of objects, employing random erasing and CutMix image augmentation techniques to enhance edge and contour information, integrating the improved convolutional block attention module (ICBAM) into the backbone structure, and replacing the sigmoid-weighted linear unit activation function with the funnel rectified linear unit activation function. Experimental results show the improved model achieving a mean average precision of 99% and notable improvements in precision. This study offers a technical solution for integrating emotion recognition into intelligent online education platforms to enhance evaluation and feedback for students.
  Keywords: algorithm efficiency optimisation; computer vision; deep neural networks; educational technology; psychology; YOLOv7.
  DOI: 10.1504/IJSNET.2024.10063953
   
  
• A Novel Approach for Real-Time Monitoring and Counting of Metro Passenger and Vehicle Flow  
  by Cong Huang, Ying Huang 
  Abstract: Accurate passenger flow information systems are key to ensuring efficient metro operations and passenger safety. To address the constraints of conventional metro passenger flow information systems, which suffer from inaccuracies in data collection and the inability to capture real-time changes in passenger flow effectively, a novel approach utilising image processing is suggested. This method monitors and accurately counts passenger and vehicle flow surrounding metro stations in real-time. This paper describes the proposed image processing method, including target detection, tracking, and counting techniques. Using data collected from high-definition cameras, this study applies advanced image processing algorithms to ensure the reliability and validity of passenger and vehicle flow monitoring results. Research has shown that the latest technique has the potential to greatly enhance the precision and tallying exactness. This enhancement can boost accuracy to over 95% when juxtaposed with older, conventional algorithms. The study not only provides a new technical solution for the metro passenger flow information system, which helps to improve the safety and efficiency of metro operation but also can provide passengers with better and more convenient traveling services and provides strong support for promoting the modernisation and intelligent development of metro operation and management.
  Keywords: metro passenger flow information system; image processing techniques; passenger flow monitoring; vehicle flow monitoring.
  DOI: 10.1504/IJSNET.2024.10064370
   
  
• Evolving Network Representation Learning Based on Recurrent Neural Network  
  by Dong-ming Chen, Mingshuo Nie, Qianqian Gan, Dongqi Wang 
  Abstract: An evolving network refers to a dynamic network with edges lasting for an extended period. Typical evolving networks include friend relationship networks and employment relationship networks. The topological structure of the evolving network remains relatively stable and exhibits learnable evolution rules, making it a hot topic for research. The primary objective of representation learning in evolving networks is to extract informative content from both the temporal and spatial dimensions and represent the network as a low-dimensional embedding vector. However, existing methods for evolving network representation learning lack time-related information. In link prediction tasks, the absence of associated information within the interval between the known interaction information and the time to be predicted by link prediction hinders the acquisition of comprehensive node representations. A novel evolving network representation learning based on recurrent neural network (ENRR) is proposed to address this problem. This algorithm leverages historical interaction information and recurrent neural network predictions to obtain network association information within the specified time interval. Comparative experiments on link prediction with baselines across multiple real-world datasets demonstrate that the proposed algorithm provides significant validity and reliability.
  Keywords: network representation learning; evolving network; recurrent neural network; link prediction; node behaviour characteristics.
  DOI: 10.1504/IJSNET.2024.10064479
   
  
• A Novel Federated Learning Approach for Routing Optimisation in Opportunistic IoT Networks  
  by Moulik Bhardwaj, Jagdeep Singh, Nitin Gupta, Kuldeep Jadon, S.K. Dhurandher 
  Abstract: Opportunistic IoT networks are the type of wireless network that operate in challenging and dynamic environments where traditional network infrastructure is unreliable, limited, or non-existent. Due to these unreliable network conditions, traditional routing algorithms can not be applied to them. Further, in today's interconnected world, where a vast amount of personal and sensitive information is transmitted over networks, it is important to address the growing concerns over the privacy and security of users' data in communication networks. To mitigate this, a Novel Federated Learning Approach for Routing Optimization in Opportunistic IoT Networks is proposed, where nodes opportunistically select the next-hop relay for message forwarding based on the current network state and local knowledge. Extensive simulation and analysis showcase the effectiveness and practicality of the proposed FLRouter in achieving efficient and privacy-aware routing within Opportunistic IoT networks. The proposed approach outperforms existing methods in delivery probability, with gains of up to 16% and 13% as buffer size increases. Additionally, it demonstrates lower overhead ratios, with reductions of up to 42% and 34% compared to existing approaches.
  Keywords: Federated Learning; Opportunistic IoT networks; Routing; Security; Privacy; ONE Simulator; Real Datasets; Keras; Tensorflow.
  DOI: 10.1504/IJSNET.2024.10064733
   
  
• Implementation of Optimisation adopted Adaptive Elliptical Curve Cryptography with Additive Homomorphic Encryption-based Privacy Preservation in Mobile Crowd Sensing Environment  
  by Domi Evangeline S, Usha G 
  Abstract: In the Location Based-Mobile Crowd Sensing (LB-MCS) network, the user has to provide their personal information to the task provider for pursuing the task. When the task provider is observed to be malicious, the task provider can monitor the user based on location and performs any kind of attack. In order to enhance the privacy preserving between the contributors, this paper aims to develop an Adaptive Elliptical Curve Cryptography with Additive Homomorphic Encryption (AECCAHE) for preserving the privacy of the user’s location information. Here, the optimisation on developed AECCAHE takes place using the Hybrid Ladybug Beetle Class Topper Optimization (HLBCTO) for handling the different location transmission of the users, and synchronisation of the contributors involving in the task in an efficient way. The experimental analysis is performed for verifying the effective privacy preservation of the developed encryption technique by comparing with conventional algorithms.
  Keywords: Location Based-Mobile Crowd Sensing; Adaptive Elliptical Curve Cryptography with Additive Homomorphic Encryption; Hybrid Ladybug Beetle Class Topper Optimization.
  DOI: 10.1504/IJSNET.2023.10064745
   
  
• Machine Learning based Fault Detection Scheme for IoT-Enabled WSNs  
  by Pravindra Shekhar Shakunt, Siba Kumar Udgata 
  Abstract: Wireless Sensor Networks (WSNs) encounter faults due to their deployment in non-deterministic and potentially hazardous environments. The process of fault identification in WSNs poses several challenges. This complexity makes it challenging to pinpoint and diagnose faults within IoT-enabled WSNs. This paper uses Extra tree and state-of-the-art machine learning classifiers to classify commonly occurring faults such as offset, drift, gain, data loss, stuck, random, and out-of-bounds at the sensor node level. First, we realistically induce the faults with different intensities to a benchmark dataset of temperature and humidity sensors. We propose sliding window-based data pre-processing techniques and various machine learning algorithms for classifying types of faults. The performance of the proposed scheme and other machine learning approaches are compared based on Specificity, Precision, Recall, Accuracy, F1-score, and AUC-ROC performance evaluation metric. Experimental study shows that our proposed scheme and the Extra tree machine learning approach are more effective than state-of-the-art approaches.
  Keywords: Wireless sensor networks; Internet of Things; Fault detection; Classification; Moving Window Average; Machine learning; Extra trees.
  DOI: 10.1504/IJSNET.2024.10064810
   
  
• An Improved Sparrow Search Algorithm and Its Application in Wireless Sensor Node Coverage Problem  
  by Jianing Guo, Yunshan Sun, Ting Liu, Yanqin Li, Teng Fei 
  Abstract: Enhancing the sparrow search algorithm (SSA) to address the wireless sensor node (WSN) coverage problem requires resolving its challenges, including inadequate search precision, susceptibility to local optima, and premature convergence. The improved algorithm demands a more efficient deployment of sensor nodes to achieve network effectiveness and cost efficiency. This paper presents an improved sparrow search algorithm (ISSA). Primarily, the finders' exponential strategy prevents the algorithm from prematurely converging to the initial point. Furthermore, ISSA integrates addition and subtraction operations to bolster joiners' strategies, facilitating comprehensive exploration of near-optimal solutions. Moreover, by integrating multiplication and division operations into finders' and scouts' strategies, ISSA randomly refreshes the solution space and reduces the impact of population initialisation. In experiments, ISSA was compared with the original SSA, four new intelligent algorithms, and two improved SSAs across 23 standard test functions. Results indicate that the improvements proposed in this paper significantly enhance optimisation accuracy and speed. Additionally, ISSA optimised WSN coverage under three distinct test parameter sets. Observed growth rates were 15.40%, 7.17%, and 17.79%, respectively. These results underscore the enhanced algorithm's superior performance in addressing the WSN coverage problem compared to the original algorithm.
  Keywords: sparrow search algorithm; SSA; arithmetic optimisation algorithm; AOA; local optima; wireless sensor node coverage; WSN.
  DOI: 10.1504/IJSNET.2024.10064943
   
  
• UAV-assisted Connectivity Recovery for Intermittent-connectivity Mobile Ad Hoc Networks  
  by Zhihui Ding, Yajie Ma 
  Abstract: Network connectivity is a fundamental requirement for the mobile ad hoc networks (MANETs). The failure of multiple nodes can disrupt the network connection, turning it into an intermittent-connectivity network. To recover the network connectivity a UAV-assisted algorithm UCRAIN is proposed to provide temporary connection solution for intermittent-connectivity network. Based on the mathematical modelling and analysis of both UAVs and network partitions, a constrained multi-objectives optimisation is developed to maximise the network coverage and minimise the size of coverage overlapping simultaneously. In order to obtain the solution of the multi-objective optimisation, a sine-cosine optimisation algorithm is adopted, which can increase the number of UAVs in a linear pattern, making it insensitive to the node density. The simulation results show that the UCRAIN algorithm can reduce the usage of UAVs by 1/2 to 2/3 comparing to other algorithms. It also has satisfactory performance in packet delivery rate, converge time and transmission delay.
  Keywords: mobile ad hoc network; MANET; intermittent-connectivity network; connectivity recovery; unmanned aerial vehicles; UAVs; network partitions.
  DOI: 10.1504/IJSNET.2024.10065334
   
  
• AROS: Human Action Recognition by Spatio-Temporal Fusion Mechanism Based on Optimized Subcarriers  
  by ZhiYong Tao, XiJun Guo, Ying Liu 
  Abstract: Wi-Fi-based human motion recognition methods are widely used for their usage and infrastructure convenience. However, the spatial diversity of MIMO causes differences in the representation of action features across antenna links. Furthermore, using all of the data is computationally time-consuming, and using a portion of the data may result in omitting crucial features. To address these issues, a human action recognition by spatio-temporal fusion mechanism based on optimised subcarriers (AROS) is proposed in this paper. Specifically, K-means adaptive clustering aims to select information-rich and complementary subcarriers with adaptive clustering cores through cluster analysis and correlation computation. The correlation-weighted fusion mechanism is presented to enhance the MIMO link characteristics. A network structure based on spatial module, temporal convolutional network, and temporal attention is presented to extract CSI spatio-temporal features. Experimental results demonstrate the effectiveness of AROS, achieving accuracies of 96.17%, 96.31%, and 94.51% in three different environments.
  Keywords: human activity recognition; HAR; channel state information; CSI; temporal convolutional network; TCN; subcarrier selection.
  DOI: 10.1504/IJSNET.2024.10065358
   
  
• Fault Prediction of Railway Track Circuit Based on Machine Learning  
  by Xin Zhang, Yan Ru 
  Abstract: Railway track circuit designs are crucial and complicated for handling movement, halting, and swapping locomotives, goods, and passengers. The circuit has intelligently automated electro-mechanical and electronic equipment to ensure locomotive and passenger safety. The unattended or prolonged failure or faults in such circuits lead to disasters; precautionary measures are hence mandatory. This article introduces a precision fault detection technique using machine learning (PFDT-ML) for equipment-specific diagnosis. The data are taken from the Railway Track Fault Detection Kaggle dataset for analysing the defective and non-defective railway tracks. This technique studies the operation synchronisation between the control and monitoring devices and their prompt functions regularly. The role of transfer learning is to retain the operational and fault states of the equipment based on their synchronisation. This learning performs function transfer for state maintenance from synchronisation to monitoring. Therefore, synchronisation alerts the faulty states to the appropriate stations for early diagnosis. On the other hand, faulty or non-functional equipment is reported before the next synchronisation interval for appropriate precautions. Thus, the learning states are swapped recurrently until the operational state of the circuit equipment is restored for synchronisation. The experimental outcome demonstrates that the recommended PFDT-ML model increases the classification accuracy ratio of 98.9%, railway track zone detection rate of 97.5%, fault prediction ratio of 96.3%, and F1-score ratio of 95.6% compared to other popular models.
  Keywords: fault prediction; railway operation; track circuit; transfer learning.
  DOI: 10.1504/IJSNET.2024.10065471
   
  
• Unsupervised Offensive Speech Detection for Multimedia based on Multilingual BERT  
  by Ge Liu, Xiaona Yang, Xiayang Shi, Yinlin Li 
  Abstract: There is a significant amount of offensive speech in multimedia, which seriously negatively impacts social stability. With the proliferation of sensor-equipped devices contributing to social media data, detecting offensive speech within this vast dataset has emerged as a critical challenge. However, most existing methods have focused only on a few high-resource languages. This paper proposes a cross-lingual aggressive transfer learning method based on bidirectional encoder representations from transformers (BERT) for automatically detecting offensive speech in low-resource languages. Initially, we utilize the multilingual BERT model to learn the characteristics of aggressive speech from a high-resource language dataset to establish an initial model. Subsequently, based on the linguistic similarity between languages, this model is transferred to low-resource languages. Experimental results demonstrate that our method achieves higher detection accuracy in multiple languages including English, Danish, Arabic, Turkish, and Greek, particularly excelling in low-resource languages.
  Keywords: natural language processing; offensive speech detection; social media.
  DOI: 10.1504/IJSNET.2024.10065475