Xiang Chen

Irrelevant gestures can cause myoelectric control systems to generate erroneous commands. To address the issue of irrelevant gesture interference, a temporal neural network based on center loss called TCNet is proposed. First, in order to mine the information between and within different periods of EMG signals, Fourier transform is used to decompose different frequency components of EMG signals, and the decomposed period is used to fold one-dimensional EMG signals in two dimensions. Second, two-dimensional convolution is used to learn the information between and within different periods of two-dimensional EMG signals, so that the model can better understand the temporal characteristics of EMG signals under different periodicities. Third, center loss is introduced to train the network, pushing features toward the center positions of the same class. This promotes small intra-class differences, facilitating better discrimination between irrelevant gestures and target gestures. Fourth, integrating Softmax output probability threshold rejection mechanism enables effective rejection of irrelevant gestures. This study focuses on 17 classes of gestures from 11 subjects, achieving a recognition rate of 91.5% for 10 target gesture classes, a rejection rate of 90.5% for 7 irrelevant gesture classes, and an average accuracy of 91.1%. This result is quite competitive compared to existing studies.

The efficient online decomposition of high-density surface electromyogram (HD-sEMG) into motor unit spike trains (MUSTs) remains a substantial challenge. In this study, a novel deep learning (DL) framework is presented. The framework leverages a hybrid DL model to capture spatiotemporal features from HD-sEMG for fast and accurate MU spike identification. Then, the continuous MUSTs are precisely reconstructed through a post-processing approach. Besides, an asymmetric loss is integrated to improve the capability of the DL model in addressing imbalanced samples. Both simulated and experimental HD-sEMG data were employed to develop and validate the framework. The offline progressive FastICA peel-off method was utilized to extract reference MUSTs to train and evaluate the proposed framework. A public HD-sEMG dataset was adopted to validate the generalizability of the framework. Results demonstrated that the proposed framework achieved accurate online estimation of MUSTs from HD-sEMG with extremely low latency (approximately 7.5[ ]ms), yielding a matching rate of [ ], [ ] and [ ] for simulated, experimental and public datasets, respectively, significantly outperforming baseline methods ([ ]). This study further enriches tools for online HD-sEMG decomposition, potentially promoting their real-time applications in fields of neurorehabilitation medicine and neural motor control.

This study aims to evaluate the crawling motor function in children with cerebral palsy (CP) using a cluster-based muscle synergy analysis scheme. Surface

Functional electrical stimulation (FES) is commonly used in clinical practice to induce muscle contractions for rehabilitation therapy. However, the collected electromyography (EMG) under FES is a mixed signal containing voluntary electromyography (VEMG), which is generated by neural system recruited motor units, and FES response, which is composed of initial spikes generated by current passage and M-waves generated by stimulation recruited motor units. To realize close-loop control of FES system or to reveal the mechanism of FES through EMG signal, it is necessary to separate the three components from mixed signal. Based on the assumption that signal sources of initial spikes and action potentials related signals (VEMG and M-waves) are independent of each other, this study presents a novel mixed FES signal decomposition algorithm termed FastICA-Comb. Specifically, in order to achieve high-quality separation of VEMG, M-waves and initial spikes, the algorithm is designed as a unique scheme combining one-stage comb filtering and two-stage fast independent component analysis (FastICA) decomposition and classification. To evaluate the proposed algorithm’s effectiveness, FES data collection experiment was conducted on 6 healthy subjects. The experimental results confirm that the FastICA-Comb algorithm has better VEMG and M-wave extraction capabilities than the classical methods including comb filter, GS-PEF, EMD-notch and blanking window in both simulated and real mixed FES signal. Therefore, the research findings provide an effective signal analysis tool for exploring the therapeutic mechanism of FES.

This paper proposes a novel approach for exercise heart rate (HR) estimation by integrating PPG quality assessment with deep neural network (DNN) modeling. A frequency-domain kurtosis (kurtF) metric is introduced to identify high-quality PPG samples, optimizing DNN training data and mitigating motion artifacts. An E-K scatter plot is used to visualize sample quality distribution, aiding dataset variability analysis. The proposed DNN model, designed for single-channel PPG input, demonstrates strong HR estimation performance on public datasets, achieving a mean absolute error (MAE) values of 3.76 bpm (PPG_DaLiA) and 3.18 bpm (IEEE-Training). Theoretical analysis and experimental validation confirm that prioritizing high-quality samples enhances model stability, accuracy, and generalizability. Additionally, a dataset quality analysis method is introduced to facilitate comparative assessments. The kurtF metric and quality-driven sample selection strategy provide a robust framework for improving HR estimation, even in data-limited scenarios. This study underscores the importance of integrating sample quality assessment into HR estimation workflows, paving the way for more accurate and reliable PPG-based HR monitoring during exercise.

Objective: This study is aimed to develop a robust myoelectric pattern recognition method for simultaneously alleviating cross-user variability and outlier motion interference. Methods: In the proposed method, a convolutional neural network (CNN)-based feature extractor is pre-trained using the data from a set of existing users. Next, a few labeled data of inlier motions recorded from a new user are utilized to implement model transfer and adaptation, while the prototype representation of each inlier motion is calibrated. In this process, a Euclidean metricbased prototypical loss is adopted to facilitate inter-class separability and intra-class compactness. Subsequently, any inlier/outlier motion is tested and identified based on a prototype matching procedure. The proposed method was evaluated on surface electromyogram signals recorded by an 8-channel armband from twenty subjects, including six inlier motions and ten outlier motions. Results: When testing with each subject following a leave-one-out testing strategy (the remaining subjects were considered to form a set of existing users for pre-training a model), the proposed method achieved average accuracies of 82.37 ± 1.21% for the inlier motion recognition and 97.21 ± 2.65% for the outlier motion rejection, respectively, and it outperformed the existing methods with statistical significance (p < 0.05). Conclusion: The proposed method yielded excellent performance in cross-user open-set myoelectric pattern recognition with only a short and simple calibration routine. Significance: Our work offers a valuable solution for improving the robustness and usability of myoelectric gestural interfaces.

Introduction: In response to the demand for a quantifiable means for assessing hand dysfunction in cerebral palsy (CP), this paper proposed and conducted a novel high-density (HD)-surface electromyography (sEMG)-based muscle force estimation framework.Methods and Results: A highly generalized source network was developed firstly based on long short-term memory (LSTM) networks and three different healthy adult HD-sEMG-force datasets, achieving a root mean square error (RMSE) of 6.31% in force estimation across various force modes; Then, transfer learning techniques were applied to fine tune the well-trained source network using data from healthy children, establishing five gesture-specific target networks that achieved RMSE below 10% in force estimation tasks independent of the subjects; Finally, a muscle force estimation experiment was conducted on 16 children with CP using the gesture-specific target networks.Conclusion: By comparing and analyzing the experimental results of CP group and healthy control group, CP children with different grades of Manual Ability Classification System (MACS), and CP children with different types of symptoms, it was verified that the abnormal EMG-force relationship obtained using the proposed muscle force estimation scheme had the potential for clinical application in the assessment of CP hand dysfunction. Muscle force estimation based on sEMG has broad application prospects in clinical practice. The research work in this paper has important value in promoting the clinical application of muscle force estimation technology based on sEMG, which is conducive to improving the quantitative assessment level of motor dysfunction.

Recently, great interests have been attracted on the online decomposition of surface electromyogram (SEMG) but current studies mainly performed validation on simulated EMG signals due to the fact that real MU activities in experimental signals were unknown. For a more comprehensive assessment of online SEMG decomposition, a two-source validation was conducted by simultaneously collecting intramuscular EMG (IEMG) and high-density SEMG signals. The IEMG signal was decomposed using a simplified version of Progressive FastICA Peel-off (PFP) method with a combination of the peel-off strategy and the valley-seeking clustering, and the decomposed motor unit (MU) spike trains were used as the ground-truth reference. For SEMG recordings, the signals within initial 5 seconds were used to offline obtain MU separation vectors and these vectors were subsequently employed to extract MU spike trains in the online stage. The matching rate of the common firing events from the ground-truth reference and online SEMG decomposition were calculated and assessed. A total of 549 and 92 MUs were identified from the SEMG and IEMG signals from 5 healthy subjects’ first dorsal interosseous muscle. All the MUs decomposed from IEMG can be matched with MUs from online SEMG decomposition and the average matching rate in the online stage was (96 ± 1) %. The results highlighted the ability of separation vectors to continuously and precisely track the same MU in the experimental SEMG signals. Our study provides a more comprehensive validation perspective of online SEMG decomposition on the experimental data.

Cross-user variability is a well-known challenge that leads to severe performance degradation and impacts the robustness of practical myoelectric control systems. To address this issue, a novel method for myoelectric recognition of finger movement patterns is proposed by incorporating a neural decoding approach with unsupervised domain adaption (UDA) learning. In our method, the neural decoding approach is implemented by extracting microscopic features characterizing individual motor unit (MU) activities obtained from a two-stage online surface electromyogram (SEMG) decomposition. A specific deep learning model is designed and initially trained using labeled data from a set of existing users. The model can update adaptively when recognizing the movement patterns of a new user. The final movement pattern was determined by a fuzzy weighted decision strategy. SEMG signals were collected from the finger extensor muscles of 15 subjects to detect seven dexterous finger-movement patterns. The proposed method achieved a movement pattern recognition accuracy of ([ ])% over seven movements under cross-user testing scenarios, much higher than that of the conventional methods using global SEMG features. Our study presents a novel robust myoelectric pattern recognition approach at a fine-grained MU level, with wide applications in neural interface and prosthesis control.