Personality reflects an individual’s enduring patterns of thought and behavior, while gait—a measurable and consistent behavioral trait—offers a unique and objective way to assess personality through natural, nonvolitional movement. Unlike traditional methods, such as self-report questionnaires, which are often subject to biases and limited accuracy, gait-based assessment provides a more direct and spontaneous measure of personality. This study introduces a gait-based personality assessment system that leverages a low-cost wearable Internet of Things (IoT) sensor to capture fine-grained motion data, including triaxial acceleration and angular velocity from the wrist and the ankle. By focusing on the natural, involuntary aspects of gait, the system avoids the biases inherent in self-presentation. Additionally, the study presents the “Gait–Personality” dataset, featuring advanced gait phase segmentation and optimized feature extraction techniques to enhance data quality. To tackle challenges like variability in stride length and cadence, a multiscale 1-D convolutional neural network (MS-1D-CNN) was developed. By utilizing convolutional layers with multiple kernel sizes, the model captures both detailed and high-level temporal features, effectively adapting to diverse gait patterns while remaining robust to sensor variability. Experimental results demonstrate classification accuracies ranging from 77% to 84.5% across the Big Five personality dimensions, validating the system’s ability to objectively capture authentic personality traits. This study establishes a reliable, cost-efficient, and scalable framework for personality assessment, offering broad implications for psychological evaluation, mental health monitoring, and human–computer interaction, with the potential for widespread real-world applications.
{"title":"Personality Assessment From Gait With Wearable IoT Sensors and Multiscale CNN","authors":"Huawei Zhang;Yu Tian;Qiaojiao Wang;Jian Li;Xiaodong Yu;Chao Lian;Gloria Jiahui Lin;Dannii Y. Yeung;Wen Jung Li;Yuliang Zhao","doi":"10.1109/JSEN.2025.3604225","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3604225","url":null,"abstract":"Personality reflects an individual’s enduring patterns of thought and behavior, while gait—a measurable and consistent behavioral trait—offers a unique and objective way to assess personality through natural, nonvolitional movement. Unlike traditional methods, such as self-report questionnaires, which are often subject to biases and limited accuracy, gait-based assessment provides a more direct and spontaneous measure of personality. This study introduces a gait-based personality assessment system that leverages a low-cost wearable Internet of Things (IoT) sensor to capture fine-grained motion data, including triaxial acceleration and angular velocity from the wrist and the ankle. By focusing on the natural, involuntary aspects of gait, the system avoids the biases inherent in self-presentation. Additionally, the study presents the “Gait–Personality” dataset, featuring advanced gait phase segmentation and optimized feature extraction techniques to enhance data quality. To tackle challenges like variability in stride length and cadence, a multiscale 1-D convolutional neural network (MS-1D-CNN) was developed. By utilizing convolutional layers with multiple kernel sizes, the model captures both detailed and high-level temporal features, effectively adapting to diverse gait patterns while remaining robust to sensor variability. Experimental results demonstrate classification accuracies ranging from 77% to 84.5% across the Big Five personality dimensions, validating the system’s ability to objectively capture authentic personality traits. This study establishes a reliable, cost-efficient, and scalable framework for personality assessment, offering broad implications for psychological evaluation, mental health monitoring, and human–computer interaction, with the potential for widespread real-world applications.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 20","pages":"39230-39245"},"PeriodicalIF":4.3,"publicationDate":"2025-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145289543","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-04DOI: 10.1109/JSEN.2025.3603629
Minho Jeon;Anil Kumar Khambampati;Seokjun Ko;Kyung Youn Kim
Toxic and corrosive by-products generated during semiconductor manufacturing can accumulate inside exhaust pipes, forming solid residues that pose risks such as pipe clogging, reduced pumping efficiency, and operational accidents. To mitigate these risks, regular preventive maintenance (PM) is required, highlighting the need for nondestructive, real-time monitoring technologies to ensure process efficiency and worker safety. This study proposes a data-driven approach to estimate the free volume index (FVI) and internal deposit conditions using electrical capacitance measurements. Two convolutional neural network (CNN) architectures 1-D convolutional neural network (1D-CNN) and 2-D convolutional neural network (2D-CNN) were developed and trained on simulated capacitance data under various deposition scenarios. The methodology was first evaluated through numerical simulations to test robustness and the performance was benchmarked against a fully connected neural network (FCNN). Subsequently, the approach was validated using real capacitance data collected from an operating semiconductor facility, thereby confirming its practical applicability. The proposed CNN-based method demonstrated high accuracy, robustness to noise, and strong generalization, offering a practical solution for early detection of clogging and process anomalies. This work contributes toward safer and efficient semiconductor manufacturing through intelligent pipe condition monitoring.
{"title":"Semiconductor Residue Deposition Monitoring in Exhaust Pipeline Based on Electrical Capacitance Tomography and Convolution Neural Network","authors":"Minho Jeon;Anil Kumar Khambampati;Seokjun Ko;Kyung Youn Kim","doi":"10.1109/JSEN.2025.3603629","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3603629","url":null,"abstract":"Toxic and corrosive by-products generated during semiconductor manufacturing can accumulate inside exhaust pipes, forming solid residues that pose risks such as pipe clogging, reduced pumping efficiency, and operational accidents. To mitigate these risks, regular preventive maintenance (PM) is required, highlighting the need for nondestructive, real-time monitoring technologies to ensure process efficiency and worker safety. This study proposes a data-driven approach to estimate the free volume index (FVI) and internal deposit conditions using electrical capacitance measurements. Two convolutional neural network (CNN) architectures 1-D convolutional neural network (1D-CNN) and 2-D convolutional neural network (2D-CNN) were developed and trained on simulated capacitance data under various deposition scenarios. The methodology was first evaluated through numerical simulations to test robustness and the performance was benchmarked against a fully connected neural network (FCNN). Subsequently, the approach was validated using real capacitance data collected from an operating semiconductor facility, thereby confirming its practical applicability. The proposed CNN-based method demonstrated high accuracy, robustness to noise, and strong generalization, offering a practical solution for early detection of clogging and process anomalies. This work contributes toward safer and efficient semiconductor manufacturing through intelligent pipe condition monitoring.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 19","pages":"37312-37326"},"PeriodicalIF":4.3,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145204548","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Visual object tracking for uncrewed aerial vehicles (UAVs) is extensively used in civil and military applications. Deep learning-based visual trackers currently dominate tracking due to their powerful modeling capabilities. However, as performance increases, so does operational latency, which limits their application on UAV platforms. To address this challenge, this study proposes a lightweight dual-template collaborative tracking framework (MCTrack), which balances both speed and accuracy in visual object tracking. Specifically, a mix-attention mechanism is proposed for feature extraction, which effectively utilizes a dual-template strategy to comprehensively extract image features. Additionally, a cross-attention mechanism is introduced to enhance target localization while reducing computational complexity. Finally, by alternately integrating these two attention mechanisms into the backbone, the model achieves target localization through the utilization of information from different semantic layers, all without introducing any additional modules. A comprehensive evaluation of five authoritative aerial benchmarks demonstrates the effectiveness of the MCTrack framework. The model achieves real-time processing at 106.1 frames per second (FPS) on an NVIDIA 2080Ti GPU. Practical testing on the NVIDIA Jetson Orin NX hardware platform achieves a speed of 33.6 FPS, confirming the practicality of MCTrack on UAV platforms.
无人机视觉目标跟踪技术在民用和军事领域有着广泛的应用。基于深度学习的视觉跟踪器由于其强大的建模能力，目前在跟踪领域占据主导地位。然而，随着性能的提高，操作延迟也在增加，这限制了它们在无人机平台上的应用。为了应对这一挑战，本研究提出了一种轻量级双模板协同跟踪框架（MCTrack），该框架平衡了视觉目标跟踪的速度和准确性。具体而言，提出了一种混合关注的特征提取机制，有效地利用双模板策略对图像特征进行综合提取。此外，引入交叉注意机制，在降低计算复杂度的同时增强目标定位。最后，通过将这两种注意机制交替集成到主干中，该模型通过利用来自不同语义层的信息实现目标定位，而无需引入任何额外的模块。对五个权威空中基准的综合评估表明了MCTrack框架的有效性。该模型在NVIDIA 2080Ti GPU上实现了每秒106.1帧（FPS）的实时处理。在NVIDIA Jetson Orin NX硬件平台上的实际测试速度达到了33.6 FPS，证实了MCTrack在无人机平台上的实用性。
{"title":"MCTrack: A New Tracker Architecture for Visual in Uncrewed Aerial Vehicles","authors":"Jiangbo He;Shengjian Guo;Ting Wang;Bo Zhao;Long Wen","doi":"10.1109/JSEN.2025.3603676","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3603676","url":null,"abstract":"Visual object tracking for uncrewed aerial vehicles (UAVs) is extensively used in civil and military applications. Deep learning-based visual trackers currently dominate tracking due to their powerful modeling capabilities. However, as performance increases, so does operational latency, which limits their application on UAV platforms. To address this challenge, this study proposes a lightweight dual-template collaborative tracking framework (MCTrack), which balances both speed and accuracy in visual object tracking. Specifically, a mix-attention mechanism is proposed for feature extraction, which effectively utilizes a dual-template strategy to comprehensively extract image features. Additionally, a cross-attention mechanism is introduced to enhance target localization while reducing computational complexity. Finally, by alternately integrating these two attention mechanisms into the backbone, the model achieves target localization through the utilization of information from different semantic layers, all without introducing any additional modules. A comprehensive evaluation of five authoritative aerial benchmarks demonstrates the effectiveness of the MCTrack framework. The model achieves real-time processing at 106.1 frames per second (FPS) on an NVIDIA 2080Ti GPU. Practical testing on the NVIDIA Jetson Orin NX hardware platform achieves a speed of 33.6 FPS, confirming the practicality of MCTrack on UAV platforms.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 20","pages":"39220-39229"},"PeriodicalIF":4.3,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145315376","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-03DOI: 10.1109/JSEN.2025.3603295
Taeyoung Kim;Yunho Jung;Seongjoo Lee
A novel lightweight hand gesture recognition approach that is based on frequency-modulated continuous-wave (FMCW) radio detection and ranging (RADAR), which aims to minimize computational complexity and memory usage as well as maintain a high recognition performance, is proposed in this article. Most of the existing methods use 2-D or 3-D features that are combined with complex neural network structures, which result in high computational costs. The proposed approach in contrast extracts four components, which include range, Doppler, azimuth, and elevation, as the 1-D time-series features. These features are fed into a neural network that comprises a multibranch temporal convolutional network (TCN), depthwise separable (DS) convolutions, and a channel attention mechanism to enhance the classification performance. The experiments were conducted with nine hand gestures that were collected from nine participants. The proposed system achieved a high accuracy of 99.38% with only 44.6 K parameters and 1.84 M floating point operations per second (FLOPs). Extensive ablation studies and comparative experiments against the existing models demonstrated that the proposed method effectively balances the performance and computational efficiency. This study validates the expressive capability of 1-D features for hand gesture recognition and suggests practical applicability in resource-constrained environments, such as embedded systems.
{"title":"Lightweight Hand Gesture Recognition Using FMCW RADAR With Multibranch Temporal Convolutional Networks and Channel Attention","authors":"Taeyoung Kim;Yunho Jung;Seongjoo Lee","doi":"10.1109/JSEN.2025.3603295","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3603295","url":null,"abstract":"A novel lightweight hand gesture recognition approach that is based on frequency-modulated continuous-wave (FMCW) radio detection and ranging (RADAR), which aims to minimize computational complexity and memory usage as well as maintain a high recognition performance, is proposed in this article. Most of the existing methods use 2-D or 3-D features that are combined with complex neural network structures, which result in high computational costs. The proposed approach in contrast extracts four components, which include range, Doppler, azimuth, and elevation, as the 1-D time-series features. These features are fed into a neural network that comprises a multibranch temporal convolutional network (TCN), depthwise separable (DS) convolutions, and a channel attention mechanism to enhance the classification performance. The experiments were conducted with nine hand gestures that were collected from nine participants. The proposed system achieved a high accuracy of 99.38% with only 44.6 K parameters and 1.84 M floating point operations per second (FLOPs). Extensive ablation studies and comparative experiments against the existing models demonstrated that the proposed method effectively balances the performance and computational efficiency. This study validates the expressive capability of 1-D features for hand gesture recognition and suggests practical applicability in resource-constrained environments, such as embedded systems.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 19","pages":"37298-37311"},"PeriodicalIF":4.3,"publicationDate":"2025-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145204534","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1109/JSEN.2025.3602594
Yu-Jie Fan;Tian-Ze Yu;Jun-Long Zhang;You-Yin Wang;Wen Bao
Obtaining a high-density temperature distribution of key components is extremely important for the safety and efficiency of thermal engines operating under extreme thermal conditions. Although the 2-D resistive sensor array (RSA) is widely used for temperature distribution measurement, its application in high-temperature environments remains to be explored. While using high-temperature-resistant materials enables the 2-D RSA to operate under such conditions, wire resistance and crosstalk from parasitic parallel paths lead to significant measurement errors. To achieve temperature distribution measurements in high-temperature environments, we propose a 2-D RSA integrated with an accurate measurement method, designed for surface deployment on high-temperature components. A $4times 4$ and an $8times 8$ 2-D RSAs, with a thickness of less than $100~mu $ m, were fabricated using screen printing, with resistance temperature detectors (RTDs) and wires made of platinum that can withstand high temperatures. Measurement errors caused by wire resistance and crosstalk are mitigated by the compensated resistance matrix approach (CRMA). The calibration of RTDs derived the temperature coefficient of resistance (TCR) and characteristic curves up to 1200 °C. Furthermore, experimental validation of the 2-D RSA confirmed its high-temperature measurement capability. The results showed that the measurements matched those of the thermal imaging camera and thermocouples with a relative error of less than 2%. This 2-D RSA is capable of accurately measuring 2-D temperature distributions in high-temperature environments up to 1200 °C.
{"title":"A 2-D Resistive Sensor Array for Temperature Distribution Measurement in High-Temperature Environments","authors":"Yu-Jie Fan;Tian-Ze Yu;Jun-Long Zhang;You-Yin Wang;Wen Bao","doi":"10.1109/JSEN.2025.3602594","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3602594","url":null,"abstract":"Obtaining a high-density temperature distribution of key components is extremely important for the safety and efficiency of thermal engines operating under extreme thermal conditions. Although the 2-D resistive sensor array (RSA) is widely used for temperature distribution measurement, its application in high-temperature environments remains to be explored. While using high-temperature-resistant materials enables the 2-D RSA to operate under such conditions, wire resistance and crosstalk from parasitic parallel paths lead to significant measurement errors. To achieve temperature distribution measurements in high-temperature environments, we propose a 2-D RSA integrated with an accurate measurement method, designed for surface deployment on high-temperature components. A <inline-formula> <tex-math>$4times 4$ </tex-math></inline-formula> and an <inline-formula> <tex-math>$8times 8$ </tex-math></inline-formula> 2-D RSAs, with a thickness of less than <inline-formula> <tex-math>$100~mu $ </tex-math></inline-formula>m, were fabricated using screen printing, with resistance temperature detectors (RTDs) and wires made of platinum that can withstand high temperatures. Measurement errors caused by wire resistance and crosstalk are mitigated by the compensated resistance matrix approach (CRMA). The calibration of RTDs derived the temperature coefficient of resistance (TCR) and characteristic curves up to 1200 °C. Furthermore, experimental validation of the 2-D RSA confirmed its high-temperature measurement capability. The results showed that the measurements matched those of the thermal imaging camera and thermocouples with a relative error of less than 2%. This 2-D RSA is capable of accurately measuring 2-D temperature distributions in high-temperature environments up to 1200 °C.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 19","pages":"37288-37297"},"PeriodicalIF":4.3,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145204557","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1109/JSEN.2025.3602873
Chenyang Guo;Haibo Yang;Guanglei Xu;Anying Chai
Linear wireless sensor networks (LWSNs) are often deployed in unsupervised scenarios such as railways. However, dynamic environments are prone to signal fading and packet loss, resulting in increased transmission delay and decreased throughput. To address this problem, this article proposes a parallel transmission DCW-CSMA/CA protocol based on the linear topology. The protocol designs a dynamic competition window adjustment algorithm based on node density to reduce multihop transmission delay. In addition, under the constraints of single-hop transmission and double-hop interference, DCW-CSMA/CA selects nodes with the highest residual energy in topology groups to perform concurrent data transmission, thereby enhancing network throughput. To ensure fair data transmission among nodes at different depths along multihop paths, this article introduces a dual-queue fair scheduling mechanism. Simulation experiments conducted on the OMNeT++ platform demonstrate that the proposed DCW-CSMA/CA protocol achieves approximately 2.3 times higher throughput than traditional CSMA protocols and exhibits lower transmission latency than three comparison protocols, thus validating its efficiency and practicality in LWSNs.
{"title":"DCW-CSMA/CA: Cross-Layer Design With Dynamic Contention Window Adaptation for Dense Wireless Sensor Networks","authors":"Chenyang Guo;Haibo Yang;Guanglei Xu;Anying Chai","doi":"10.1109/JSEN.2025.3602873","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3602873","url":null,"abstract":"Linear wireless sensor networks (LWSNs) are often deployed in unsupervised scenarios such as railways. However, dynamic environments are prone to signal fading and packet loss, resulting in increased transmission delay and decreased throughput. To address this problem, this article proposes a parallel transmission DCW-CSMA/CA protocol based on the linear topology. The protocol designs a dynamic competition window adjustment algorithm based on node density to reduce multihop transmission delay. In addition, under the constraints of single-hop transmission and double-hop interference, DCW-CSMA/CA selects nodes with the highest residual energy in topology groups to perform concurrent data transmission, thereby enhancing network throughput. To ensure fair data transmission among nodes at different depths along multihop paths, this article introduces a dual-queue fair scheduling mechanism. Simulation experiments conducted on the OMNeT++ platform demonstrate that the proposed DCW-CSMA/CA protocol achieves approximately 2.3 times higher throughput than traditional CSMA protocols and exhibits lower transmission latency than three comparison protocols, thus validating its efficiency and practicality in LWSNs.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 19","pages":"37462-37471"},"PeriodicalIF":4.3,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145204541","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Self-supervised monocular depth estimation realizes training without depth labeling data by mining the geometric consistency of image sequences, which has important application value in fields, such as autonomous driving. Traditional methods rely on complex CNN and transformer hybrid architectures to balance local and global features but face problems, such as a large number of model parameters and low computational efficiency, which severely limit the deployment capability of edge devices. Although the existing lightweight methods reduce the number of parameters through techniques, such as depth-separable convolution and channel compression, there are still have problems, such as insufficient multiscale feature fusion, limited interaction ability of global and local context information, and loss of details at the edge of the depth map. To solve these problems, we propose LM-DualNet, a novel architecture with dual-path attention enhancement. Specifically, the encoder integrates a dynamic local context-aware (DLCA) module for capturing fine-grained local structures, and a dual-axis gated attention (DAGA) module that constructs two parallel attention paths-spatial and channel-to jointly model positional dependencies and cross-channel correlations. In the decoder, we design a multiscale depth enhancement (MSDE) module to refine edge regions and enhance depth continuity. Experiments on the KITTI dataset show that the absolute relative error and squared relative error of LM-DualNet have decreased to 0.106 and 0.731, respectively, and the accuracy has reached 88.8%, which is a good improvement compared with other state-of-the-art algorithms.
{"title":"Dual-Path Attention Network for Lightweight Self-Supervised Monocular Depth Estimation","authors":"Chao Zhang;Tian Tian;Cheng Han;Tiancheng Shao;Mi Zhou;Shichao Zhao","doi":"10.1109/JSEN.2025.3601212","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3601212","url":null,"abstract":"Self-supervised monocular depth estimation realizes training without depth labeling data by mining the geometric consistency of image sequences, which has important application value in fields, such as autonomous driving. Traditional methods rely on complex CNN and transformer hybrid architectures to balance local and global features but face problems, such as a large number of model parameters and low computational efficiency, which severely limit the deployment capability of edge devices. Although the existing lightweight methods reduce the number of parameters through techniques, such as depth-separable convolution and channel compression, there are still have problems, such as insufficient multiscale feature fusion, limited interaction ability of global and local context information, and loss of details at the edge of the depth map. To solve these problems, we propose LM-DualNet, a novel architecture with dual-path attention enhancement. Specifically, the encoder integrates a dynamic local context-aware (DLCA) module for capturing fine-grained local structures, and a dual-axis gated attention (DAGA) module that constructs two parallel attention paths-spatial and channel-to jointly model positional dependencies and cross-channel correlations. In the decoder, we design a multiscale depth enhancement (MSDE) module to refine edge regions and enhance depth continuity. Experiments on the KITTI dataset show that the absolute relative error and squared relative error of LM-DualNet have decreased to 0.106 and 0.731, respectively, and the accuracy has reached 88.8%, which is a good improvement compared with other state-of-the-art algorithms.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 19","pages":"37419-37428"},"PeriodicalIF":4.3,"publicationDate":"2025-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145204481","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-29DOI: 10.1109/JSEN.2025.3597711
Yue Xu;Baoquan Hu;Quan Pan
To address the issues of long-term dependency and insufficient local feature extraction in traditional methods when processing hypersonic target trajectory data, this article proposes an innovative trajectory prediction method that integrates equivariant graph neural networks (EGNNs) and Transformer architecture. Specifically, by constructing dynamic graph structures to model the geometric motion characteristics of the target, EGNN uses an equivariant message-passing mechanism to extract spatial features with SE (3) covariance. Meanwhile, the Transformer, with its multihead attention mechanism and geometric correction attention module, explicitly captures the long-term spatiotemporal dependencies in the trajectory data. To further enhance the model’s performance, an improved whale optimization algorithm (IWOA) is proposed, which dynamically regulates the learning rate using Lyapunov stability theory and combines Hamiltonian dynamics to reconstruct the predation strategy, significantly improving global search ability and convergence efficiency. Additionally, the AdamW optimizer is used to independently handle the weight decay term, effectively suppressing overfitting. The experimental results show that the proposed method achieves a position prediction root-mean-square error (RMSE) of 532.1 m and a velocity prediction RMSE of 268.3 m/s on the Northwestern Polytechnical University (NPU) trajectory dataset, improving accuracy by 23.8% and 38.8%, respectively, compared to the next-best method. Moreover, the model’s parameter count (2.75 M) and computational cost (5.68 GFLOPs) are significantly lower than those of the comparison models. Ablation experiments verify the effectiveness of the EGNN equivariant feature, IWOA dynamic optimization mechanism, and AdamW regularization strategy, providing a solution that balances both accuracy and efficiency for hypersonic target trajectory prediction.
{"title":"A Hypersonic Target Trajectory Prediction Method Based on EGNN and Transformer","authors":"Yue Xu;Baoquan Hu;Quan Pan","doi":"10.1109/JSEN.2025.3597711","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3597711","url":null,"abstract":"To address the issues of long-term dependency and insufficient local feature extraction in traditional methods when processing hypersonic target trajectory data, this article proposes an innovative trajectory prediction method that integrates equivariant graph neural networks (EGNNs) and Transformer architecture. Specifically, by constructing dynamic graph structures to model the geometric motion characteristics of the target, EGNN uses an equivariant message-passing mechanism to extract spatial features with SE (3) covariance. Meanwhile, the Transformer, with its multihead attention mechanism and geometric correction attention module, explicitly captures the long-term spatiotemporal dependencies in the trajectory data. To further enhance the model’s performance, an improved whale optimization algorithm (IWOA) is proposed, which dynamically regulates the learning rate using Lyapunov stability theory and combines Hamiltonian dynamics to reconstruct the predation strategy, significantly improving global search ability and convergence efficiency. Additionally, the AdamW optimizer is used to independently handle the weight decay term, effectively suppressing overfitting. The experimental results show that the proposed method achieves a position prediction root-mean-square error (RMSE) of 532.1 m and a velocity prediction RMSE of 268.3 m/s on the Northwestern Polytechnical University (NPU) trajectory dataset, improving accuracy by 23.8% and 38.8%, respectively, compared to the next-best method. Moreover, the model’s parameter count (2.75 M) and computational cost (5.68 GFLOPs) are significantly lower than those of the comparison models. Ablation experiments verify the effectiveness of the EGNN equivariant feature, IWOA dynamic optimization mechanism, and AdamW regularization strategy, providing a solution that balances both accuracy and efficiency for hypersonic target trajectory prediction.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 19","pages":"37499-37511"},"PeriodicalIF":4.3,"publicationDate":"2025-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145204540","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
One of the fundamental challenges in wireless sensor networks (WSNs) is ensuring reliable data transmission while optimizing energy efficiency. This article addresses this challenge by proposing a novel multicriteria cluster head (CH) selection algorithm and an adaptive relay strategy for energy-harvesting WSNs in smart city environments. Our approach integrates four key metrics for CH selection: proximity to the sink, residual energy (RE), transmission reliability, and power consumption. Additionally, we introduce a dynamic multihop routing protocol to mitigate obstacles and enhance network reliability. The proposed work is formulated as a nonconvex optimization problem and transformed into a convex problem for efficient solution. Simulation results demonstrate significant improvements compared to previous works: energy efficiency increases by up to 45% over entire rounds, the network transmission ratio improves packet delivery rates by 30%–40%, and outage probability is reduced to near-zero levels under stable network conditions. These metrics are evaluated under varying modified subsistence ratios (0.15 and 0.45), highlighting the robustness of our method. These improvements originate from multicriteria CH selection, obstacle-aware routing, and energy management, collectively extending network lifetime and reliability for large-scale smart city deployments.
{"title":"An Energy-Aware Cluster Head Selection and Relay Strategy for Efficient Data Transmission in Smart City WSNs","authors":"Sajjad Nouri;Faranak Reyhani;Javad Musevi Niya;Behzad Mozaffari Tazehkand","doi":"10.1109/JSEN.2025.3602354","DOIUrl":"https://doi.org/10.1109/JSEN.2025.3602354","url":null,"abstract":"One of the fundamental challenges in wireless sensor networks (WSNs) is ensuring reliable data transmission while optimizing energy efficiency. This article addresses this challenge by proposing a novel multicriteria cluster head (CH) selection algorithm and an adaptive relay strategy for energy-harvesting WSNs in smart city environments. Our approach integrates four key metrics for CH selection: proximity to the sink, residual energy (RE), transmission reliability, and power consumption. Additionally, we introduce a dynamic multihop routing protocol to mitigate obstacles and enhance network reliability. The proposed work is formulated as a nonconvex optimization problem and transformed into a convex problem for efficient solution. Simulation results demonstrate significant improvements compared to previous works: energy efficiency increases by up to 45% over entire rounds, the network transmission ratio improves packet delivery rates by 30%–40%, and outage probability is reduced to near-zero levels under stable network conditions. These metrics are evaluated under varying modified subsistence ratios (0.15 and 0.45), highlighting the robustness of our method. These improvements originate from multicriteria CH selection, obstacle-aware routing, and energy management, collectively extending network lifetime and reliability for large-scale smart city deployments.","PeriodicalId":447,"journal":{"name":"IEEE Sensors Journal","volume":"25 19","pages":"37445-37461"},"PeriodicalIF":4.3,"publicationDate":"2025-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145204536","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","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：