Pub Date : 2025-01-27DOI: 10.1109/ACCESS.2025.3534662
Salah Ud Din;Shah Khusro;Farman Ali Khan;Munir Ahmad;Oualid Ali;Taher M. Ghazal
Offensive language is a type of unacceptable language that is impolite amongst individuals, specific community groups, and society as well. With the advent of various social media platforms, offensive language usage has been widely reported, thus developing a toxic online environment that has real-life endangers within society. Therefore, to foster a culture of respect and acceptance, a prompt response is needed to combat offensive content. On the other hand, the identification of offensive language has become a challenging task, specifically in low-resource languages such as Urdu. Urdu text poses challenges because of its unique features, complex script, and rich morphology. Applying methods directly that work in other languages is difficult. It also requires exploring new linguistic features and computational techniques on a relatively large dataset to ensure the results can be generalized effectively. Unfortunately, the Urdu language got very limited attention from the research community due to the scarcity of language resources and the non-availability of high-quality datasets and models. This study addresses those challenges, firstly by collecting and annotating a dataset of 12020 Urdu tweets using OLID taxonomy as a benchmark. Secondly, by extracting character-level and word-level features based on bag-of-words, n-grams and TFIDF representation. Finally, an extensive series of experiments were conducted on the extracted features using seven machine learning classifiers to identify the most effective features and classifiers. The experimental findings indicate that word unigrams, character trigrams, and word TFIDF are the most prominent ones. Similarly, among the classifiers, logistic regression and support vector machine attained the highest accuracy of 86% and F1-Score of 75%.
{"title":"An Automatic Approach for the Identification of Offensive Language in Perso-Arabic Urdu Language: Dataset Creation and Evaluation","authors":"Salah Ud Din;Shah Khusro;Farman Ali Khan;Munir Ahmad;Oualid Ali;Taher M. Ghazal","doi":"10.1109/ACCESS.2025.3534662","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3534662","url":null,"abstract":"Offensive language is a type of unacceptable language that is impolite amongst individuals, specific community groups, and society as well. With the advent of various social media platforms, offensive language usage has been widely reported, thus developing a toxic online environment that has real-life endangers within society. Therefore, to foster a culture of respect and acceptance, a prompt response is needed to combat offensive content. On the other hand, the identification of offensive language has become a challenging task, specifically in low-resource languages such as Urdu. Urdu text poses challenges because of its unique features, complex script, and rich morphology. Applying methods directly that work in other languages is difficult. It also requires exploring new linguistic features and computational techniques on a relatively large dataset to ensure the results can be generalized effectively. Unfortunately, the Urdu language got very limited attention from the research community due to the scarcity of language resources and the non-availability of high-quality datasets and models. This study addresses those challenges, firstly by collecting and annotating a dataset of 12020 Urdu tweets using OLID taxonomy as a benchmark. Secondly, by extracting character-level and word-level features based on bag-of-words, n-grams and TFIDF representation. Finally, an extensive series of experiments were conducted on the extracted features using seven machine learning classifiers to identify the most effective features and classifiers. The experimental findings indicate that word unigrams, character trigrams, and word TFIDF are the most prominent ones. Similarly, among the classifiers, logistic regression and support vector machine attained the highest accuracy of 86% and F1-Score of 75%.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"19755-19769"},"PeriodicalIF":3.4,"publicationDate":"2025-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10854428","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105994","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Emergence in complex systems is often compounded by diverse information and rapid technological acceleration. The problems and behaviors of increasingly complex systems continue confining practitioners’ systems engineering capabilities to maintain performance consistency. While systems engineering provides a process for integrating various engineering disciplines to deliver desired end results, systems thinking (ST) provides the mechanism for drawing broad perspectives on the configuration, patterns, and cycles of complex systems, with a view to analyzing and improving system performance. ST can, therefore, be construed as an essential skill for designing and managing complex systems that need to sustain desired specifications. Although several methods exist in the extant literature to appraise the ST skills of practitioners, none has been recommended for prediction and diagnostic purposes. To fill this void, this research study aims to develop and validate a Bayesian network tool that incorporates seven main factors and the corresponding underpinning sub-factors that influence individual ST skills, as identified by Jaradat and Keating. The study seeks to answer whether differences in systems thinking skills are evident between practitioners in two sectors, namely defense and industry/business. The results indicate that all the main ST factors are imperative to predicting overall individual ST skills for defense and industry/business practitioners. However, defense practitioners scored higher along six dimensions, resulting in a higher overall individual ST score than industry practitioners. Although industry practitioners scored higher than defense practitioners on the independence (autonomy) dimension, this dimension alone was insufficient to strengthen the overall ST skills above that of defense practitioners.
{"title":"Predicting Individual Systems Thinking Skills Using Bayesian Network Approach","authors":"Niamat Ullah Ibne Hossain;Raed Jaradat;Morteza Nagahi;Alex Gorod","doi":"10.1109/ACCESS.2025.3534632","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3534632","url":null,"abstract":"Emergence in complex systems is often compounded by diverse information and rapid technological acceleration. The problems and behaviors of increasingly complex systems continue confining practitioners’ systems engineering capabilities to maintain performance consistency. While systems engineering provides a process for integrating various engineering disciplines to deliver desired end results, systems thinking (ST) provides the mechanism for drawing broad perspectives on the configuration, patterns, and cycles of complex systems, with a view to analyzing and improving system performance. ST can, therefore, be construed as an essential skill for designing and managing complex systems that need to sustain desired specifications. Although several methods exist in the extant literature to appraise the ST skills of practitioners, none has been recommended for prediction and diagnostic purposes. To fill this void, this research study aims to develop and validate a Bayesian network tool that incorporates seven main factors and the corresponding underpinning sub-factors that influence individual ST skills, as identified by Jaradat and Keating. The study seeks to answer whether differences in systems thinking skills are evident between practitioners in two sectors, namely defense and industry/business. The results indicate that all the main ST factors are imperative to predicting overall individual ST skills for defense and industry/business practitioners. However, defense practitioners scored higher along six dimensions, resulting in a higher overall individual ST score than industry practitioners. Although industry practitioners scored higher than defense practitioners on the independence (autonomy) dimension, this dimension alone was insufficient to strengthen the overall ST skills above that of defense practitioners.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"20133-20148"},"PeriodicalIF":3.4,"publicationDate":"2025-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10854422","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106021","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-27DOI: 10.1109/ACCESS.2025.3534405
Danil D. Kazimirov;Ekaterina O. Rybakova;Vitalii V. Gulevskii;Arseniy P. Terekhin;Elena E. Limonova;Dmitry P. Nikolaev
The Hough (discrete Radon) transform (HT/DRT) is a digital image processing tool that has become indispensable in many application areas, ranging from general image processing to neural networks and X-ray computed tomography. The utilization of the HT in applied problems demands its computational efficiency and increased accuracy. The de facto standard algorithm for the fast HT is the Brady-Yong algorithm. However, it is applicable only to images of a power-of-two width. The algorithm that generalizes the Brady-Yong algorithm for non-power-of-two image width with the same asymptotic complexity is known, but it has not been studied neither in terms of the constant in the asymptotics nor in accuracy. Thus, supported both by theory and experiments, generalization of the Brady-Yong algorithm for images of arbitrary size while maintaining asymptotic computational complexity and acceptable accuracy remains of paramount necessity. In this paper, we proposed 5 novel algorithms that incorporate the core idea of the Brady-Yong algorithm and are suitable for computing the fast HT for images of arbitrary size. We investigated the properties of 5 new algorithms, along with one previously proposed algorithm from the literature, through both theoretical analysis and experimental validation. As one of our major contributions, among the proposed algorithms, we singled out a $FHT2DT$ algorithm and proved that it provides a substantial compromise between accuracy and computational complexity. The $FHT2DT$ algorithm is significantly more accurate than the algorithm previously suggested in the literature and, hence, $FHT2DT$ can substitute it in practical applications. During the analysis of the properties of the proposed algorithms, we created a map that characterizes the algorithms based on their accuracy and speed parameters. Users can select the method that best suits their needs — whether prioritizing computational complexity or accuracy — by referring to our map.
{"title":"Generalizing the Brady-Yong Algorithm: Efficient Fast Hough Transform for Arbitrary Image Sizes","authors":"Danil D. Kazimirov;Ekaterina O. Rybakova;Vitalii V. Gulevskii;Arseniy P. Terekhin;Elena E. Limonova;Dmitry P. Nikolaev","doi":"10.1109/ACCESS.2025.3534405","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3534405","url":null,"abstract":"The Hough (discrete Radon) transform (HT/DRT) is a digital image processing tool that has become indispensable in many application areas, ranging from general image processing to neural networks and X-ray computed tomography. The utilization of the HT in applied problems demands its computational efficiency and increased accuracy. The de facto standard algorithm for the fast HT is the Brady-Yong algorithm. However, it is applicable only to images of a power-of-two width. The algorithm that generalizes the Brady-Yong algorithm for non-power-of-two image width with the same asymptotic complexity is known, but it has not been studied neither in terms of the constant in the asymptotics nor in accuracy. Thus, supported both by theory and experiments, generalization of the Brady-Yong algorithm for images of arbitrary size while maintaining asymptotic computational complexity and acceptable accuracy remains of paramount necessity. In this paper, we proposed 5 novel algorithms that incorporate the core idea of the Brady-Yong algorithm and are suitable for computing the fast HT for images of arbitrary size. We investigated the properties of 5 new algorithms, along with one previously proposed algorithm from the literature, through both theoretical analysis and experimental validation. As one of our major contributions, among the proposed algorithms, we singled out a <inline-formula> <tex-math>$FHT2DT$ </tex-math></inline-formula> algorithm and proved that it provides a substantial compromise between accuracy and computational complexity. The <inline-formula> <tex-math>$FHT2DT$ </tex-math></inline-formula> algorithm is significantly more accurate than the algorithm previously suggested in the literature and, hence, <inline-formula> <tex-math>$FHT2DT$ </tex-math></inline-formula> can substitute it in practical applications. During the analysis of the properties of the proposed algorithms, we created a map that characterizes the algorithms based on their accuracy and speed parameters. Users can select the method that best suits their needs — whether prioritizing computational complexity or accuracy — by referring to our map.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"20101-20132"},"PeriodicalIF":3.4,"publicationDate":"2025-01-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10854475","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106158","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-24DOI: 10.1109/ACCESS.2025.3534024
Ji Li;Chenyi Liao
Accurate identification of tea leaf diseases is crucial for intelligent tea cultivation and monitoring. However, the complex environment of tea plantations—affected by weather variations and uneven lighting—poses significant challenges for building effective disease recognition models using raw field-captured images. To address this, we propose a method that combines two-stage image segmentation with an improved conditional generative adversarial network (IC-GAN). The two-stage segmentation approach, integrating graph cuts and support vector machines (SVM), effectively isolates disease regions from complex backgrounds. The IC-GAN augments the dataset by generating high-quality synthetic disease images for model training. Finally, an Inception Embedded Pooling Convolutional Neural Network (IDCNN) is developed for disease recognition. Experimental results demonstrate that the segmentation method improves recognition accuracy from 53.36% to 75.63%, while the IC-GAN increases the dataset size. The IDCNN achieves 97.66% accuracy, 97.36% recall, and a 96.98% F1 score across three types of tea diseases. Comparative evaluations on two additional datasets further confirm the method’s robustness and accuracy, offering a practical solution to reduce tea production losses and improve quality.
{"title":"Tea Disease Recognition Based on Image Segmentation and Data Augmentation","authors":"Ji Li;Chenyi Liao","doi":"10.1109/ACCESS.2025.3534024","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3534024","url":null,"abstract":"Accurate identification of tea leaf diseases is crucial for intelligent tea cultivation and monitoring. However, the complex environment of tea plantations—affected by weather variations and uneven lighting—poses significant challenges for building effective disease recognition models using raw field-captured images. To address this, we propose a method that combines two-stage image segmentation with an improved conditional generative adversarial network (IC-GAN). The two-stage segmentation approach, integrating graph cuts and support vector machines (SVM), effectively isolates disease regions from complex backgrounds. The IC-GAN augments the dataset by generating high-quality synthetic disease images for model training. Finally, an Inception Embedded Pooling Convolutional Neural Network (IDCNN) is developed for disease recognition. Experimental results demonstrate that the segmentation method improves recognition accuracy from 53.36% to 75.63%, while the IC-GAN increases the dataset size. The IDCNN achieves 97.66% accuracy, 97.36% recall, and a 96.98% F1 score across three types of tea diseases. Comparative evaluations on two additional datasets further confirm the method’s robustness and accuracy, offering a practical solution to reduce tea production losses and improve quality.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"19664-19677"},"PeriodicalIF":3.4,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10852315","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105511","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-24DOI: 10.1109/ACCESS.2025.3533537
Shuqin Wang;Na Cheng;Yan Hu
Addressing the challenges of limited accuracy in anomaly detection within comprehensive environmental monitoring of industrial and mining enterprises, and the constraints posed by singular data modalities, this study proposes an integration of a multimodal Long Short-Term Memory (LSTM) model with the Contrastive Language-Image Pretraining (CLIP) model. The initial phase employs ResNet within the CLIP model for extracting image features, and a Transformer for encoding text features. Subsequently, feature vectors obtained from monitoring images and text are fused using a rudimentary concatenation method to generate a joint embedding representation. Principal Component Analysis (PCA) is then applied to diminish the dimensionality of the amalgamated features derived from environmental monitoring images, descriptive texts, and sensor data collected by industrial and mining enterprises. Finally, a multimodal LSTM model is leveraged to detect anomalies in the monitoring data by capturing long-term dependencies within time series information. The model was trained and evaluated using real-time data from a coal mining enterprise’s environmental monitoring system spanning March to September 2023. Results reveal that the multimodal LSTM-CLIP model achieved an anomaly detection accuracy of 0.98 in environmental monitoring, marking a 0.10 improvement over the unimodal LSTM model, with a response time of merely 110.25 milliseconds. These findings underscore the efficacy of the multimodal LSTM-CLIP model in integrating multimodal information, thereby significantly enhancing the accuracy of anomaly detection and the speed of environmental anomaly warnings, ultimately ensuring the safety of industrial and mining enterprises.
{"title":"Comprehensive Environmental Monitoring System for Industrial and Mining Enterprises Using Multimodal Deep Learning and CLIP Model","authors":"Shuqin Wang;Na Cheng;Yan Hu","doi":"10.1109/ACCESS.2025.3533537","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533537","url":null,"abstract":"Addressing the challenges of limited accuracy in anomaly detection within comprehensive environmental monitoring of industrial and mining enterprises, and the constraints posed by singular data modalities, this study proposes an integration of a multimodal Long Short-Term Memory (LSTM) model with the Contrastive Language-Image Pretraining (CLIP) model. The initial phase employs ResNet within the CLIP model for extracting image features, and a Transformer for encoding text features. Subsequently, feature vectors obtained from monitoring images and text are fused using a rudimentary concatenation method to generate a joint embedding representation. Principal Component Analysis (PCA) is then applied to diminish the dimensionality of the amalgamated features derived from environmental monitoring images, descriptive texts, and sensor data collected by industrial and mining enterprises. Finally, a multimodal LSTM model is leveraged to detect anomalies in the monitoring data by capturing long-term dependencies within time series information. The model was trained and evaluated using real-time data from a coal mining enterprise’s environmental monitoring system spanning March to September 2023. Results reveal that the multimodal LSTM-CLIP model achieved an anomaly detection accuracy of 0.98 in environmental monitoring, marking a 0.10 improvement over the unimodal LSTM model, with a response time of merely 110.25 milliseconds. These findings underscore the efficacy of the multimodal LSTM-CLIP model in integrating multimodal information, thereby significantly enhancing the accuracy of anomaly detection and the speed of environmental anomaly warnings, ultimately ensuring the safety of industrial and mining enterprises.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"19964-19978"},"PeriodicalIF":3.4,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10852209","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106187","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-24DOI: 10.1109/ACCESS.2025.3533702
Mitra Nabian Dehaghani;Tarmo Korõtko;Argo Rosin
The integration of distributed generation (DG), renewable energy sources (RES), and power electronic converters into distribution systems (DSs) has introduced significant power quality (PQ) challenges, such as voltage fluctuations, harmonic distortions, and transients. These issues can undermine the reliability and stability of power systems, making it essential to address them to ensure a consistent and resilient power supply, especially as RES adoption continues to grow. While previous reviews have explored artificial intelligence (AI) applications for PQ management, most have been limited to specific AI techniques or targeted PQ problems, such as harmonics. This review, however, offers a comprehensive synthesis of AI-based approaches across a wide range of PQ applications, encompassing detection, classification, and improvement, while also considering the specific PQ issues addressed in each case. By adopting an integrated approach, this review identifies key research gaps, particularly the limited focus on leveraging AI to control power converters in RESs for PQ improvement, as most existing studies emphasize devices like active power filters, compensators, and conditioners. The review also evaluates the effectiveness of these AI methods in terms of accuracy and the extent of total harmonic distortion (THD) reduction. In addition, it provides novel insights that can help guide researchers, engineers, and industry professionals toward developing more adaptive, scalable, and robust PQ solutions. Finally, future research directions are proposed to advance AI-based PQ management, facilitating the integration of AI into diverse and evolving power systems.
{"title":"AI Applications for Power Quality Issues in Distribution Systems: A Systematic Review","authors":"Mitra Nabian Dehaghani;Tarmo Korõtko;Argo Rosin","doi":"10.1109/ACCESS.2025.3533702","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533702","url":null,"abstract":"The integration of distributed generation (DG), renewable energy sources (RES), and power electronic converters into distribution systems (DSs) has introduced significant power quality (PQ) challenges, such as voltage fluctuations, harmonic distortions, and transients. These issues can undermine the reliability and stability of power systems, making it essential to address them to ensure a consistent and resilient power supply, especially as RES adoption continues to grow. While previous reviews have explored artificial intelligence (AI) applications for PQ management, most have been limited to specific AI techniques or targeted PQ problems, such as harmonics. This review, however, offers a comprehensive synthesis of AI-based approaches across a wide range of PQ applications, encompassing detection, classification, and improvement, while also considering the specific PQ issues addressed in each case. By adopting an integrated approach, this review identifies key research gaps, particularly the limited focus on leveraging AI to control power converters in RESs for PQ improvement, as most existing studies emphasize devices like active power filters, compensators, and conditioners. The review also evaluates the effectiveness of these AI methods in terms of accuracy and the extent of total harmonic distortion (THD) reduction. In addition, it provides novel insights that can help guide researchers, engineers, and industry professionals toward developing more adaptive, scalable, and robust PQ solutions. Finally, future research directions are proposed to advance AI-based PQ management, facilitating the integration of AI into diverse and evolving power systems.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"18346-18365"},"PeriodicalIF":3.4,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10852279","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105462","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the continuous technological progress, especially the development in biometrics, gait recognition has shown broad application prospects in healthcare (e.g., health monitoring), security (e.g., assisted identity verification), and human-computer interaction. However, individual differences, such as changes in physical condition, and environmental variability, such as differences in lighting, can impact its accuracy. Based on the information derived from the gait contour sequence during walking (such as temporal and spatial information), this study proposes an improved gait recognition method based on the GaitSet model, which improves video-based gait recognition performance by combining gait energy images and silhouette images to form a multimodal representation. The experimental results showed a significant performance improvement compared with the original model, especially in subjects with bags. Large-sample training experiment results based on the CASIA-B database indicated that the recognition rates in the Normal (NM), Bag (BG), and Coat (CL) states were 95.8%, 89.3%, and 72.5%, respectively, and that in the CL state achieved a significant improvement of 3.3%.
{"title":"Research on Gait Recognition Based on GaitSet and Multimodal Fusion","authors":"Xiling Shi;Wenqiang Zhao;Huandou Pei;Hongru Zhai;Yongxia Gao","doi":"10.1109/ACCESS.2025.3533571","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533571","url":null,"abstract":"With the continuous technological progress, especially the development in biometrics, gait recognition has shown broad application prospects in healthcare (e.g., health monitoring), security (e.g., assisted identity verification), and human-computer interaction. However, individual differences, such as changes in physical condition, and environmental variability, such as differences in lighting, can impact its accuracy. Based on the information derived from the gait contour sequence during walking (such as temporal and spatial information), this study proposes an improved gait recognition method based on the GaitSet model, which improves video-based gait recognition performance by combining gait energy images and silhouette images to form a multimodal representation. The experimental results showed a significant performance improvement compared with the original model, especially in subjects with bags. Large-sample training experiment results based on the CASIA-B database indicated that the recognition rates in the Normal (NM), Bag (BG), and Coat (CL) states were 95.8%, 89.3%, and 72.5%, respectively, and that in the CL state achieved a significant improvement of 3.3%.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"20017-20024"},"PeriodicalIF":3.4,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10852208","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105757","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-24DOI: 10.1109/ACCESS.2025.3533555
Sanyi Li;Hongyu Guan;Peng Liu;Weichao Yue;Qian Wang
Stochastic Configuration Networks (SCNs) perform well in machine learning and data mining tasks in complex data environments. However, traditional SCNs have limitations in network size and computation time. To address these issues, this paper proposes an improved version of SCNs. There are two key improvements: First, the stopping condition for generating neurons is optimized to improve the effectiveness of new neurons. Second, the regularization parameter r is adjusted dynamically to speed up the learning process. These improvements are trying to increase the efficiency of SCN construction, reduce the number of redundant neurons, and shorten the overall computation time. Experiments comparing this method with existing ones show that the proposed approach not only reduces network complexity but also effectively decreases training time. In addition, experimental results using baseline datasets and UCI databases show that the number of nodes required for oscn is reduced by approximately 50% and the computation time is reduced by approximately 40% compared to traditional algorithms.
{"title":"An Improved Stochastic Configuration Networks With Compact Structure and Parameter Adaptation","authors":"Sanyi Li;Hongyu Guan;Peng Liu;Weichao Yue;Qian Wang","doi":"10.1109/ACCESS.2025.3533555","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533555","url":null,"abstract":"Stochastic Configuration Networks (SCNs) perform well in machine learning and data mining tasks in complex data environments. However, traditional SCNs have limitations in network size and computation time. To address these issues, this paper proposes an improved version of SCNs. There are two key improvements: First, the stopping condition for generating neurons is optimized to improve the effectiveness of new neurons. Second, the regularization parameter r is adjusted dynamically to speed up the learning process. These improvements are trying to increase the efficiency of SCN construction, reduce the number of redundant neurons, and shorten the overall computation time. Experiments comparing this method with existing ones show that the proposed approach not only reduces network complexity but also effectively decreases training time. In addition, experimental results using baseline datasets and UCI databases show that the number of nodes required for oscn is reduced by approximately 50% and the computation time is reduced by approximately 40% compared to traditional algorithms.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"18141-18151"},"PeriodicalIF":3.4,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10852165","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105613","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-24DOI: 10.1109/ACCESS.2025.3533703
K. Velu;N. Jaisankar
Parkinson’s Disease (PD) is a chronic and progressive neurological disorder that impairs the body’s nervous system pathways. This disruption results in multiple complications related to movement and control, manifesting as symptoms such as tremors, rigidity, and impaired coordination. In the initial phases of PD, individuals have trouble with speech and exhibit a slow rate of verbal expression. Dysphonia is seventy to ninety percent of persons with Parkinson’s disease report a speech impairment or modification in speech, and it serves as a preliminary indicator of the disease. Consequently, speech can be a crucial modality in the initial phase of Parkinson’s disease prediction. In literature, diverse Machine Learning models are employed for Parkinson’s disease diagnosis utilizing voice analysis. Challenges such as class imbalance, feature selection, and interpretable predictive analysis still need to be addressed. Furthermore, the precision and reliability of the predictive outcomes are crucial to enhance healthcare services. Consequently, we propose an Explainable balanced Recursive Feature Importance with Logistic Regression (XRFILR) model to address the abovementioned issues. The proposed model extracts the pertinent features using an RFE with a Logistic Regression classifier and evaluates the feature significance in Parkinson’s disease prediction using eXplainable Artificial Intelligence. We employed the seven machine learning classifiers the model offers to diagnose Parkinson’s disease using significant speech data. Among these ML models, the proposed model achieved an accuracy of 96.46%, surpassing existing machine learning techniques.
{"title":"Design of an Early Prediction Model for Parkinson’s Disease Using Machine Learning","authors":"K. Velu;N. Jaisankar","doi":"10.1109/ACCESS.2025.3533703","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533703","url":null,"abstract":"Parkinson’s Disease (PD) is a chronic and progressive neurological disorder that impairs the body’s nervous system pathways. This disruption results in multiple complications related to movement and control, manifesting as symptoms such as tremors, rigidity, and impaired coordination. In the initial phases of PD, individuals have trouble with speech and exhibit a slow rate of verbal expression. Dysphonia is seventy to ninety percent of persons with Parkinson’s disease report a speech impairment or modification in speech, and it serves as a preliminary indicator of the disease. Consequently, speech can be a crucial modality in the initial phase of Parkinson’s disease prediction. In literature, diverse Machine Learning models are employed for Parkinson’s disease diagnosis utilizing voice analysis. Challenges such as class imbalance, feature selection, and interpretable predictive analysis still need to be addressed. Furthermore, the precision and reliability of the predictive outcomes are crucial to enhance healthcare services. Consequently, we propose an Explainable balanced Recursive Feature Importance with Logistic Regression (XRFILR) model to address the abovementioned issues. The proposed model extracts the pertinent features using an RFE with a Logistic Regression classifier and evaluates the feature significance in Parkinson’s disease prediction using eXplainable Artificial Intelligence. We employed the seven machine learning classifiers the model offers to diagnose Parkinson’s disease using significant speech data. Among these ML models, the proposed model achieved an accuracy of 96.46%, surpassing existing machine learning techniques.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"17457-17472"},"PeriodicalIF":3.4,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10852288","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105879","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-24DOI: 10.1109/ACCESS.2025.3534138
Ryan Aalund;Vincent Philip Paglioni
Embedded Systems are used in extreme conditions, often for long lifespans; as such, ensuring hardware reliability is essential. Additionally, the applications of embedded systems can be safety-critical or costly in the event of a failure. Applications and environments such as these demand high reliability. This literature survey explores the challenges of achieving hardware reliability in embedded systems. It examines critical works using different methodologies and viewpoints to summarize hardware reliability comprehensively. The paper discusses the main failure modes identified in embedded systems hardware, evaluates various mitigation strategies, and identifies emerging trends influencing the future of embedded system design. By critically analyzing existing literature, this survey is a resource for future research efforts focused on growing the reliability of embedded systems. Finally, this paper outlines the motivation and first methods for a systems-level approach to embedded systems reliability.
{"title":"Enhancing Reliability in Embedded Systems Hardware: A Literature Survey","authors":"Ryan Aalund;Vincent Philip Paglioni","doi":"10.1109/ACCESS.2025.3534138","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3534138","url":null,"abstract":"Embedded Systems are used in extreme conditions, often for long lifespans; as such, ensuring hardware reliability is essential. Additionally, the applications of embedded systems can be safety-critical or costly in the event of a failure. Applications and environments such as these demand high reliability. This literature survey explores the challenges of achieving hardware reliability in embedded systems. It examines critical works using different methodologies and viewpoints to summarize hardware reliability comprehensively. The paper discusses the main failure modes identified in embedded systems hardware, evaluates various mitigation strategies, and identifies emerging trends influencing the future of embedded system design. By critically analyzing existing literature, this survey is a resource for future research efforts focused on growing the reliability of embedded systems. Finally, this paper outlines the motivation and first methods for a systems-level approach to embedded systems reliability.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"17285-17302"},"PeriodicalIF":3.4,"publicationDate":"2025-01-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10852318","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105991","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：