International Journal of Intelligent Systems Technologies and Applications
Forthcoming articles have been peer-reviewed and accepted for publication but are pending final changes, are not yet published and may not appear here in their final order of publication until they are assigned to issues. Therefore, the content conforms to our standards but the presentation (e.g. typesetting and proof-reading) is not necessarily up to the Inderscience standard. Additionally, titles, authors, abstracts and keywords may change before publication. Articles will not be published until the final proofs are validated by their authors.
Forthcoming articles must be purchased for the purposes of research, teaching and private study only. These articles can be cited using the expression "in press". For example: Smith, J. (in press). Article Title. Journal Title.
Articles marked with this shopping trolley icon are available for purchase - click on the icon to send an email request to purchase.
Online First articles are published online here, before they appear in a journal issue. Online First articles are fully citeable, complete with a DOI. They can be cited, read, and downloaded. Online First articles are published as Open Access (OA) articles to make the latest research available as early as possible.
Articles marked with this Open Access icon are Online First articles. They are freely available and openly accessible to all without any restriction except the ones stated in their respective CC licenses.
International Journal of Intelligent Systems Technologies and Applications (7 papers in press)
Performance Evaluation of Entropy Based Cuboid Selection Technique For Human Interaction Recognition by Aditi Jahagirdar Abstract: Local feature-based approaches are proved most successful for the application of human action recognition. This work aims to study the effect of the number of cuboids chosen on the performance of human action recognition. A 3D region, called cuboid, is extracted around every spatio-temporal interest point and an entropy-based cuboid selection method is implemented for choosing the cuboids having maximum information. The proposed method is evaluated using UT interaction and SBU kinect interaction datasets. Results show that maximum classification accuracy and F1Score are obtained when the top 40 cuboids are selected for the feature extraction. The classification accuracy and F1Score increase when the number of selected cuboids is increased from 10 to 40 and remain constant when the number is increased beyond 40. The selected 40 cuboids are then used for feature extraction. To carry out this experiment, wavelet coefficients are extracted from selected cuboids are used as features. Keywords: cuboids; discrete wavelet transform; entropy; human action recognition; interaction detection; spatio temporal interest points. DOI: 10.1504/IJISTA.2022.10049131
Local Directional Gradients Extension for Recognizing Face and Facial Expressions by Ayache Farid, Alti Adel Abstract: This paper proposes new descriptors for recognizing face and facial expressions. Our descriptors consist in combining local directional gradients with SVM (Support Vector Machine) linear classification. This combination allows extracting discriminant facial expressions features for better classification accuracy with good efficiency than existing classifiers. Both descriptors are built based on the reduced texture features extracted from the face based on magnitude and orientation maps on the horizontal and vertical coordinates. JAFFE and YALE benchmarks have been used to evaluate the accuracy and execution time of the requested face in the classification process. The experimental results are very promising and show that the proposed descriptors are effective and efficient compared to current works. Keywords: Local Descriptor; Histogram of Directional Gradient; Texture Feature Analysis; SVM Classifier; Facial Expression Recognition; Histogram of Directional Gradient Generalized; Face recognition.
Deep learning approach for classifying ischemic stroke using DWI sequences of brain MRIs by Sukanta Sabut, Prasanta Patra, Arun Ray Abstract: Stroke is an emergency condition and must be treated immediately to increase the survivability rate. The diagnosis of stroke by clinicians is rely on the analysis of magnetic resonance imaging (MRI) images of brain. The present standard approach to identify the brain stroke is done manually by the radiologist which is time consuming and operator dependability. The computer-aided diagnosis is effective and having better detection accuracy. We propose an automatic detection approach to identify the ischemic stroke infarcts based on deep neural network (DNN) architecture using diffusion-weighted imaging (DWI) sequences of brain MRIs. A total 192 stroke affected MRI images of the brain were collected at Kalinga Institute of Medical Science, India. Initially all the images were pre-processed to reduce the noise and then taken for segmentation of stroke infarct using a Delaunay triangulation (DT) approach. The DT approach is a fully automated process that extract binary mask of the object without initial estimation of the number of clusters. Thirty-four important features are extracted from the segmented infarct lesions and then classified with the DNN classifier. From the evaluation results, we achieved high detection rate with sensitivity 89.18%, specificity 95.37, Jaccard Index 81.46% and accuracy of 92.8% in classifying the ischemic stroke into three sub-types (LACS, lacunar syndrome; PLACS, partial anterior circulation syndrome; TACS, total anterior circulation stroke). The obtained results is compared with few published articles. It has been observed that the deep learning approach is an effective way to detect the stroke infarcts in clinical practices using brain images. Keywords: Computer-aided detection; image segmentation; Delaunay triangulation; classification; deep neural network.
Breast Cancer Detection by Fusion of Deep Features with CNN Extracted Features by Liang Zhou, Amita Nandal, Todor Ganchev, Arvind Dhaka Abstract: Breast cancer growth has become a typical factor now-a-days. Physician experience of diagnosing and detecting breast cancer can be assisted by using some computerized features extraction and classification algorithms. In the recent times breast cancer can be diagnosed by classifying tumors. In this paper breast cancer identification and analysis is done by using machine learning statistical analysis. The proposed technique has proven to improve the exactness of foreseeing predicting cancer. The proposed method used optimized recording condition of the input image and later introduces a new interpretable feature for the identification. The simulation results are compared with conventional methods by using accuracy, sensitivity and specificity for performance assessment of the identification process. Keywords: Convolutional Neural Networks; Segmentation; Tumors; Machine Learning Algorithm; Classification; Image Processing.
Integration of Deep Learning Techniques for Sentiment and Emotion Analysis of Social Media Data by H.S. Hota, Dinesh Sharma, Nilesh Verma Abstract: : Sentiment Analysis (SA) and Emotion Analysis (EA) are commonly used to understand people's feelings and opinions on a given topic. COVID-19 is an emerging infectious disease that is rapidly spreading around the world. The mental state of a country's population is more or less the same worldwide. Machine Learning (ML) techniques are commonly utilized to analyze human sentiments and emotions. Two popular Deep Learning (DL) techniques: Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM), are being applied in several areas. In this study, we propose a hybrid of CNN and LSTM to improve the performance of the classification model. The two different models, the Sentiment Analysis Model (SAM) and the Emotional Analysis Model (EAM), were developed using benchmark data, which produces 91.11% and 89.39% accuracy, respectively, by integrating CNN and LSTM. Integration of two or more techniques significantly improves performance by utilizing both techniques. The results of the experiments demonstrate that the proposed hybrid technique outperforms other individual DL techniques. Keywords: Convolutional Neural Network (CNN); Sentiment Analysis (SA); Emotion Analysis (EA); COVID-19; Deep Learning (DL); Long Short-Term Memory (LSTM).
Access Selection in Heterogeneous Wireless Networks Based on User Preferences by Jamal Haydar, Abed Ellatif Samhat, Guy Pujolle Abstract: Access selection is an important key in heterogeneous networks, and the design of a new algorithm for decision is not a trivial task. Different aspects must be taken into consideration while designing a new decision algorithm, including both users requirements (in terms of resources, QoS, users preferences), and operator policies that aim to maximize the utilization of its network capacity and to deliver services with acceptable QoS levels for the largest number of customers. Thus, in this paper we propose a new selection algorithm based on users preferences. The comparison between the proposed scenarios is given based on several performance indicators. The results show the improvement achieved by increasing the resource utilization and therefore the overall system capacity. Keywords: access selection; heterogeneous networks; user preferences; QoS; resource utilization.
Fake Face Detection in Video using Shallow Deep Learning Architectures by Hai Thanh Nguyen, Tinh Cong Dao, Thao Minh Nguyen Phan, Tai Tan Phan Abstract: Deep learning techniques have been used in various disciplines, ranging from simple data processing to complicated image classification tasks. Deepfakes is a deep learning approach with benefits and drawbacks impacting the world. However, deepfakes are now cutting-edge technology being exploited for nefarious purposes such as the breach of human privacy and identity. Because deep learning is advancing rapidly daily, people use AI to produce deepfakes videos and images. Hence newer AI technology to detect deepfakes is critical. Therefore, the study has proposed detecting video and image deepfakes based on convolutional neural network (CNN) combined with the long short-term memory (LSTM) model, which constructs a deep learning model classifying images and video deepfakes. The proposed model investigated a novel approach to research more powerful models which can be applied to any large dataset. The experimental results demonstrated that the proposed method had achieved promising performance on modified datasets from Celeb-DF with high AUC performance up to 0.7584 and MCC reaching 0.558. Besides, this paper presents brief research on creating and detecting the image and video deepfakes technologies and points out the challenges of using deepfakes in many different contexts. Keywords: deepfakes; deep learning; convolutional neural network; CNN; long short-term memory; LSTM; deepfake detection. DOI: 10.1504/IJISTA.2022.10050825