{"title":"A Real-Time Artifact Removal System for Closed-Loop Deep-Brain Stimulation","authors":"Chenghao Xing;Xi Cheng;Hao Feng;Bin Wu;Yingnan Nie;Chunguang Chu;Xin Zhang;Qiyu Niu;Jia Xiu;Bowen Geng;Liang Chen;Shouyan Wang","doi":"10.1109/TNSRE.2025.3597916","DOIUrl":"10.1109/TNSRE.2025.3597916","url":null,"abstract":"This paper presents a novel real-time signal processing method for removing local field potential (LFP) artifacts during deep-brain stimulation (DBS). Real-time artifact removal is essential for closed-loop DBS systems, as they rely on real-time, artifact-free LFPs to provide stimulation feedback. Building on previous stimulation-sampling synchronization methods, this work introduces a dynamic template subtraction method that achieves precise and efficient real-time removal of stimulation artifacts. By leveraging stimulation-sampling synchronization, the method enables real-time template alignment and artifact removal through subtraction. It can operate even at low sampling rates, requiring a minimum of twice the stimulation frequency. The artifact templates are dynamically updated to adapt to changes in stimulation artifacts, ensuring robust and accurate performance over time. The method was evaluated through simulations and in vitro and in vivo experiments. Simulation tests validated its theoretical feasibility, while it successfully removed stimulation artifacts in vitro, relative errors in the power spectral density between the recovered and reference LFPs in the examined frequency band (1—150 Hz) were 0.64%, 0.31%, 0.58%, and 0.73% under stimulation at 20, 60, 90, and 130 Hz, respectively. In vivo, the method successfully recorded artifact-free LFPs in real time and supported beta-triggered closed-loop DBS. In an additional in vivo evaluation using a commercial medical device, the method recorded artifact-free LFPs with a sampling rate of 260 Hz (twice the stimulation frequency of 130 Hz). The proposed artifact removal method provides important technical support for realizing lightweight closed-loop DBS systems.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3237-3245"},"PeriodicalIF":5.2,"publicationDate":"2025-08-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11126183","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144859079","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Lingyun Yan;Haohua Xiu;Yuyang Wei;Kaitian Cao;Xudong Luo;Yiqi Li
{"title":"Reducing Muscular Effort in Exoskeletons: A Bioinspired Design and Hierarchical Motion Recognition Framework","authors":"Lingyun Yan;Haohua Xiu;Yuyang Wei;Kaitian Cao;Xudong Luo;Yiqi Li","doi":"10.1109/TNSRE.2025.3599383","DOIUrl":"10.1109/TNSRE.2025.3599383","url":null,"abstract":"This study aims to address the limitations of traditional exoskeleton designs by developing a biomimetic actuation path and a hierarchical motion recognition framework to improve integration with human biomechanics and reduce muscular effort during walking. Methods: A musculoskeletal model was used to quantify lower limb muscle force patterns, enabling the design of actuation paths aligned with natural muscle contraction trajectories. A hierarchical motion recognition system, combining an auto-encoder and an artificial neural network (ANN), was developed for real-time identification of gait events, activity levels, and walking speeds. Two biomechanical-inspired control strategies were implemented to replicate natural movement patterns and adapt to dynamic forces during walking. Results: Experimental validation through EMG-based walking trials demonstrated a significant reduction in muscle activity. Specifically, the exoskeleton reduced the maximum voluntary isometric contraction (%MVIC) of the soleus muscle by 12.39% and the gastrocnemius by 12.32% compared to unassisted walking. Conclusion: The proposed design effectively integrates human-exoskeleton interaction, reduces muscular effort, and provides precise motion assistance, offering a novel approach to incorporating muscle force analysis in wearable robotics. Significance: This work advances the field of exoskeleton technology by introducing a quantitative biomechanical approach for actuation path optimization and real-time motion recognition, with potential applications in rehabilitation, assistive devices, and human locomotion enhancement.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3212-3224"},"PeriodicalIF":5.2,"publicationDate":"2025-08-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11126533","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144859080","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jamie A. O’Reilly;Hassapong Sunthornwiriya-Amon;Naradith Aparprasith;Pannapa Kittichalao;Pornnaphas Chairojwong;Thanabodee Klai-On;Edward W. Lannon
{"title":"Blind Source Separation of Event-Related Potentials Using Recurrent Neural Network","authors":"Jamie A. O’Reilly;Hassapong Sunthornwiriya-Amon;Naradith Aparprasith;Pannapa Kittichalao;Pornnaphas Chairojwong;Thanabodee Klai-On;Edward W. Lannon","doi":"10.1109/TNSRE.2025.3598795","DOIUrl":"10.1109/TNSRE.2025.3598795","url":null,"abstract":"Event-related potentials (ERPs) are a superposition of electric potential differences generated by neurophysiological activity associated with psychophysiological events. Spatiotemporal dissociation of underlying signal sources can supplement conventional ERP analysis and improve source localization. However, sources separated by independent component analysis (ICA) can be challenging to interpret because of redundant or illusory components and indeterminant polarity and scale. Hence, we have developed a recurrent neural network (RNN) method for blind source separation. The RNN transforms input step pulse signals representing events into corresponding ERP difference waveforms. Source waveforms are obtained from penultimate layer units and scalp maps are obtained from feed-forward output layer weights that project these source waveforms onto EEG electrode amplitudes. An interpretable, sparse source representation is achieved by incorporating L1 regularization of signals obtained from the penultimate layer of the network during training. This RNN method was applied to four ERP difference waveforms (MMN, N170, N400, P3) from the open-access ERP CORE database, and ICA was applied to the same data for comparison. The RNN decomposed real ERPs into eleven spatially and temporally separate sources that were less noisy, tended to be more ERP-specific, and were less similar to each other than ICA-derived sources. The RNN sources also had less ambiguity between source waveform amplitude, scalp potential polarity, and equivalent current dipole orientation than ICA sources. In conclusion, the proposed RNN blind source separation method can be effectively applied to average ERP waves and holds promise for further development as a computational model of event-related neural signals.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3271-3280"},"PeriodicalIF":5.2,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11124940","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144855132","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yun Qin;Siwei Xiong;Congyu Xu;Wei Zhao;Jiaxin Xie;Fali Li;Yunfang Li;Dezhong Yao;Tiejun Liu
{"title":"Effects of Cardiac Activity on Somatosensory Evoked High-Frequency Oscillations","authors":"Yun Qin;Siwei Xiong;Congyu Xu;Wei Zhao;Jiaxin Xie;Fali Li;Yunfang Li;Dezhong Yao;Tiejun Liu","doi":"10.1109/TNSRE.2025.3598757","DOIUrl":"10.1109/TNSRE.2025.3598757","url":null,"abstract":"Recent neuroscience research has shed light on heart-brain interactions during diverse information processes across perception, affective, and cognitive domains. It remains unclear how the heartbeat-related interoceptive pathway affects the neural responses of somatosensory information processing. In this study, we combined EEG, ECG, and DTI to examine the effect of heart-brain interaction on cortical somatosensory processing and investigated both the cardiac phase and heart rate effects on somatosensory-evoked high-frequency oscillations (HFOs). First, we examined the somatosensory cortex activity in terms of HFOs along the cardiac cycle and observed an attenuated HFO response in the systole phase. Moreover, voluntary hyperventilation (VH) was adopted as the approach to interoceptive exposure, and a significant HFO decrease was observed after VH in both the systole and diastole phases. Then, we constructed robust fusion models to demonstrate the combined effects of cardiac activity, brain structure, and somatosensory stimulation input on the HFO response. The results showed that there was an important predictive effect of heart rate on somatosensory neural oscillations. These findings revealed the essential regulatory effect of dynamic cardiac activity on brain response during somatosensory information processing, and they may provide a better understanding of the mechanisms underlying the heart-brain interaction by integrating interoceptive and exteroceptive signals.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3293-3302"},"PeriodicalIF":5.2,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11124931","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144855133","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Joris Gentinetta;Michael F. Fernandez;Junqing Qiao;Maria Ramos Gonzalez;Hugh M. Herr
{"title":"Biophysical Models With Adaptive Online Learning for Direct Neural Control of Prostheses","authors":"Joris Gentinetta;Michael F. Fernandez;Junqing Qiao;Maria Ramos Gonzalez;Hugh M. Herr","doi":"10.1109/TNSRE.2025.3599114","DOIUrl":"10.1109/TNSRE.2025.3599114","url":null,"abstract":"Direct neural control of multi-articulating prosthetic hands is critical for achieving dexterous manipulation in unstructured environments. However, such control — predicting continuous movements over independent degrees of freedom — remains confined to research settings. In contrast, pattern recognition systems are widely employed for their simple, user-friendly training procedures, though their limitation to a set of discrete whole-hand poses restricts functionality. To bridge this gap, we designed a direct neural controller and a training procedure to support adaptive retraining, enabling users to improve controller predictions or incorporate new movements using a single RGB camera. It explicitly models musculoskeletal dynamics and employs a neural network-based method for motor intent disambiguation, which we term “synergy inversion”. The defined dynamics constrain the predicted kinetics and kinematics to a physiologically realizable manifold, while synergy inversion can capture nonlinear patterns of muscle coactivation missing from traditional musculoskeletal models. In experiments with eight biologically intact participants and two individuals with unilateral transradial amputation, the proposed paradigm predicted trajectories for seven degrees of freedom and improved performance through online learning, achieving lower error than both purely neural and purely biophysical baseline models. This work represents a step toward the adoption of direct neural control of upper extremity prostheses in real-world settings, offering the flexibility of pattern recognition training within a more performant control framework.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3201-3211"},"PeriodicalIF":5.2,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11124939","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144855131","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Caleb J. Thomson;W. Caden Hamrick;Jakob W. Travis;Michael D. Adkins;Patrick P. Maitre;Steven R. Edgely;Jacob A. George
{"title":"Simultaneous and Proportional Myoelectric Control of Multiple Degrees of Freedom in Individuals With Chronic Hemiparesis","authors":"Caleb J. Thomson;W. Caden Hamrick;Jakob W. Travis;Michael D. Adkins;Patrick P. Maitre;Steven R. Edgely;Jacob A. George","doi":"10.1109/TNSRE.2025.3599062","DOIUrl":"10.1109/TNSRE.2025.3599062","url":null,"abstract":"Stroke is a leading cause of disability worldwide, with most survivors experiencing chronic motor deficits. Myoelectric orthoses, controlled by residual muscle activity from the paretic limb, can restore upper-limb function to patients. However, existing commercial myoelectric orthoses are limited to only a single hand motion with fixed force output. In the adjacent field of myoelectric prostheses, regression algorithms have enabled simultaneous and proportional position control over multiple degrees of freedom (DOFs), which in turn has improved user dexterity. Here, we explore, for the first time, the ability to regress the kinematic position of multiple DOFs in parallel from paretic muscle activity using a Kalman filter. We collected data from seven hemiparetic patients and systematically explored the root mean squared error (RMSE) of kinematic predictions for various degrees of freedom. We show that proportional position control is possible for multiple hand and wrist motions and that unidirectional DOFs perform better than bidirectional DOFs. Using previously reported RMSEs from healthy participants as a benchmark, we found that 86% of hemiparetic patients achieved functional 2-DOF control, 57% achieved functional 3-DOF control, and 29% achieved functional 4-DOF control. Performance was similar across patient characteristics and different combinations of DOFs. This work demonstrates that multi-DOF regression is readily achievable for some hemiparetic patients. Restoring wrist motion, in addition to grasping, could have a substantial impact on the dexterity and independence of hemiparetic patients. As such, this work serves as an important first step towards multi-DOF assistive upper-limb exoskeletons.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3246-3258"},"PeriodicalIF":5.2,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11124928","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144855135","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Javier Navallas;Lucia Biurrun;Cristina Mariscal;Silvia Recalde-Villamayor;Armando Malanda;Javier Rodríguez-Falces
{"title":"Motor Neuron Loss Detection Based on EMG Probability Density Function Shape Descriptors","authors":"Javier Navallas;Lucia Biurrun;Cristina Mariscal;Silvia Recalde-Villamayor;Armando Malanda;Javier Rodríguez-Falces","doi":"10.1109/TNSRE.2025.3599103","DOIUrl":"10.1109/TNSRE.2025.3599103","url":null,"abstract":"EMG interference pattern analysis is routinely used in the assessment of motor neuron loss. We propose systematizing interference pattern analysis by recording an isometric ramp contraction of a muscle, from minimum to maximum activation level. Three EMG probability density function (PDF) shape descriptors are then employed to quantify the PDF evolution assessing EMG filling through contraction: filling factor, negentropy, and kurtosis. The three filling curves are fitted with an exponential model, and the decay constant parameters are employed to obtain a feature vector that characterizes the EMG filling behavior of the muscle. Results show a tendency of the filling curves to shorten and not reach saturation when neuropathy is simulated, and a subsequent dependency of the decay constant parameters with neuropathy progression. We demonstrate, with a set of real signals and through simulation experiments, the ability of the features to be used by a classification system to detect motor neuron loss. With the set of real signals (from 40 subjects with L5 radiculopathy and 40 healthy controls), results show a 0.86 sensibility and 0.84 specificity, indicating a promising performance when incorporated into clinical decision support systems.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3189-3200"},"PeriodicalIF":5.2,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11124927","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144855134","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Lingyun Gao;Rui Zhang;Mengru Xu;Yi Sun;Gang Li;Yu Sun
{"title":"Brain Network Analysis Reveals Age-Related Differences in Topological Reorganization During Vigilance Decline","authors":"Lingyun Gao;Rui Zhang;Mengru Xu;Yi Sun;Gang Li;Yu Sun","doi":"10.1109/TNSRE.2025.3598197","DOIUrl":"10.1109/TNSRE.2025.3598197","url":null,"abstract":"To mitigate the economic losses and safety risks caused by reduced alertness of individuals in the context of an aging workforce, mental fatigue among the elderly is an issue worthy of in-depth exploration. Despite convergent studies on cognitive aging, the differential alterations in brain network topology between the elderly and young individuals during vigilance decline remain unclear. Here, a prolonged 30-min psychomotor vigilance task (PVT) was employed to induce mental fatigue, where both behavioral performance and electroencephalography (EEG) data were collected from healthy elderly (n =30) and young participants (n =40). Subsequently, EEG functional connectivity was constructed and the differences in network topological properties between the two groups were quantitatively evaluated based on global and nodal metrics. Both groups an exhibited age-independent significant decline in behavioral performance with time on task. Moreover, age-related dysconnectivity pattern was revealed over a wide frequency range (<inline-formula> <tex-math>$1-45$ </tex-math></inline-formula> Hz) in the elderly group, which further developed toward less optimal network architecture. Specifically, significant deficits in nodal efficiency were revealed in most of the brain regions, and the frontal area exhibited significant age-by-time interaction effect, which was attributed to a significant decline in the elderly group. Statistically significant correlation between behavioral and network metrics was also found. Overall, our results provide some of the first quantitative insights for revealing the neural mechanisms of age differences during mental fatigue, which may contribute to the rational arrangement of personnel in real-world scenarios with high alertness demands.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3160-3170"},"PeriodicalIF":5.2,"publicationDate":"2025-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11123522","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144835002","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Foot Pressure-Based Abnormal Gait Recognition With Multi-Scale Cross-Attention Fusion","authors":"Menghao Yuan;Yan Wang;Xiaohu Zhou;Meijiang Gui;Aihui Wang;Chen Wang;Guotao Li;Hongnian Yu;Lin Meng;Zengguang Hou","doi":"10.1109/TNSRE.2025.3597639","DOIUrl":"10.1109/TNSRE.2025.3597639","url":null,"abstract":"Abnormal gait recognition plays a critical role in healthcare, particularly for the early diagnosis and continuous monitoring of neurological and musculoskeletal disorders, such as Parkinson’s disease and orthopedic injuries. This study proposes MSCAF-Gait, a Multi-Scale Cross-Attention Fusion Network designed specifically for abnormal gait recognition using foot pressure sensors. MSCAF-Gait incorporates multi-scale convolutional modules with channel and spatial attention mechanisms to effectively capture features across temporal, channel, and spatial dimensions. A novel cross-attention fusion module further enhances feature representation, enabling precise recognition of diverse abnormal gait patterns. To facilitate this research, we introduce the Pressure-Insole Abnormal Gait (PIAG) dataset, comprising gait data associated with common neurological and musculoskeletal abnormalities. Extensive experiments on the publicly available Gait in Parkinson’s Disease (GaitinPD) dataset and our self-constructed PIAG dataset validate the effectiveness of MSCAF-Gait. Specifically, the model achieves 99.61% accuracy in Parkinsonian gait recognition and 98.88% accuracy in Parkinson’s severity classification. On the PIAG dataset, which includes multiple abnormal gait patterns, MSCAF-Gait attains a high accuracy of 99.42%. Notably, these results are obtained with a lightweight architecture characterized by reduced FLOPs and parameter count, demonstrating that MSCAF-Gait offers both high accuracy and computational efficiency, making it well-suited for real-time deployment on wearable platforms.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3146-3159"},"PeriodicalIF":5.2,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11122534","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144821372","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mingjie Yan;Zhe Chen;Jianmin Li;Jinhua Li;Lizhi Pan
{"title":"Simultaneous Estimation of Wrist Joint Angle and Torque During Isokinetic Contraction Based on HD-sEMG","authors":"Mingjie Yan;Zhe Chen;Jianmin Li;Jinhua Li;Lizhi Pan","doi":"10.1109/TNSRE.2025.3596839","DOIUrl":"10.1109/TNSRE.2025.3596839","url":null,"abstract":"The establishment of a natural and smooth human-computer interface is crucial for myoelectric control, which requires an effective decoding method for movement intention. Based on high-density surface electromyography (HD-sEMG), this study explored a method to simultaneously estimate wrist joint angle and torque during isokinetic contraction. Ten able-bodied individuals were instructed to complete wrist isokinetic flexion and extension tasks with different movement patterns, and the HD-sEMG signals were collected. To decode these signals, a convolutional neural network (CNN) incorporating the global attention mechanism was established, named global attention convolutional neural network (GACNN). Six other decoding models were also used to continuously estimate the wrist joint angle and torque, including support vector machine (SVM), residual network (ResNet), long short-term memory (LSTM), transformer-based model (TBM), muscle synergy-based graph attention networks (MSGAT-LSTM), and spatio-temporal feature extraction network (STFEN). Evaluation metrics including normalized root mean square error (NRMSE) and Pearson’s correlation coefficient (PCC) were applied to evaluate the estimation performance of the seven models. The GACNN showed significantly better estimation performance than SVM, LSTM, ResNet, STFEN and it also demonstrated superior performance over TBM and MSGAT-LSTM in some estimation cases. On average, for all subjects, NRMSE and PCC of the GACNN were <inline-formula> <tex-math>$0.080~pm ~0.013$ </tex-math></inline-formula> and <inline-formula> <tex-math>$0.955~pm ~0.016$ </tex-math></inline-formula>. The result shows the superiority of the neural network incorporating global attention mechanism, which is of great significance for the application of human-computer interaction.","PeriodicalId":13419,"journal":{"name":"IEEE Transactions on Neural Systems and Rehabilitation Engineering","volume":"33 ","pages":"3137-3145"},"PeriodicalIF":5.2,"publicationDate":"2025-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11119647","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144798973","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}