This paper presents an effective transfer learning (TL) strategy for the realization of surface electromyography (sEMG)-based gesture recognition with high generalization and low training burden. To realize the idea of taking a well-trained model as the feature extractor of the target networks, a convolutional neural network (CNN)-based source network is designed and trained as the general gesture EMG feature extraction network firstly. To fully cover possible muscle activation modes related to hand gestures, 30 hand gestures involving various states of finger joints, elbow joint and wrist joint are selected to compose the source task. Then, two types of target networks, in the forms of CNN-only and CNN+LSTM (long short-term memory) respectively, are designed with the same CNN architecture as the feature extraction network. Finally, gesture recognition experiments on three different target gesture datasets are carried out under TL and Non-TL strategies respectively. The experimental results verify the validity of the proposed TL strategy in improving hand gesture recognition accuracy and reducing training burden. For both the CNN-only and the CNN+LSTM target networks, on the three target datasets from new users, new gestures and different collection scheme, the proposed TL strategy improves the recognition accuracy by 10%~38%, reduces the training time to tens of times, and guarantees the recognition accuracy of more than 90% when only 2 repetitions of each gesture are used to fine-tune the parameters of target networks. The proposed TL strategy has important application value for promoting the development of myoelectric control systems.

Electroencephalography (EEG) has been widely used for studying brain function. As cortical signals recorded by the EEG are very weak, they are often obscured by motion artifacts and various non-brain physiological activities, such as eye movement, heart rhythms, and muscle activity, adversely affecting subsequent analysis and interpretation. Over the past decades, a number of techniques have been developed for preprocessing EEG recordings to improve the signal-to-noise ratio. However, based on our extensive literature survey, despite an increasing trend recently, only 14.72% of published studies on EEG preprocessing were involved with the removal of electromygraphic (EMG) artifacts. Given that ambulatory healthcare systems are continuously emerging, artifacts induced by muscle contraction become unavoidable, whereas in the past data tended to be collected in well-controlled clinical/laboratory settings. Motivated by the fact that EMG artifact removal is becoming an important issue to be addressed, we investigated the state-of-the-art muscle artifact removal methods systematically and comparatively, from the perspective of signal processing. In this review, we first present the signal characteristics from brain and muscle activity, and highlight the importance of this issue for subsequent artifact removal. We then provide an overview and taxonomy of representative methods. Based on the results in reported studies, we describe the pros and cons of different methods and give suggestions on selecting a suitable technique in different scenarios (e.g., single-channel, few-channel, and multichannel). Finally, we discuss remaining challenges and provide feasible recommendations for further exploration in this field.

Automatic arrhythmia detection using 12-lead electrocardiogram (ECG) signal plays a critical role in early prevention and diagnosis of cardiovascular diseases. In the previous studies on automatic arrhythmia detection, most methods concatenated 12 leads of ECG into a matrix, and then input the matrix to a variety of feature extractors or deep neural networks for extracting useful information. Under such frameworks, these methods had the ability to extract comprehensive features (known as integrity) of 12-lead ECG since the information of each lead interacts with each other during training. However, the diverse lead-specific features (known as diversity) among 12 leads were neglected, causing inadequate information learning for 12-lead ECG. To maximize the information learning of multi-lead ECG, the information fusion of comprehensive features with integrity and lead-specific features with diversity should be taken into account. In this paper, we propose a novel Multi-Lead-Branch Fusion Network (MLBF-Net) architecture for arrhythmia classification by integrating multi-loss optimization to jointly learning diversity and integrity of multi-lead ECG. MLBF-Net is composed of three components: 1) multiple lead-specific branches for learning the diversity of multi-lead ECG; 2) cross-lead features fusion by concatenating the output feature maps of all branches for learning the integrity of multi-lead ECG; 3) multi-loss co-optimization for all the individual branches and the concatenated network. We demonstrate our MLBF-Net on China Physiological Signal Challenge 2018 which is an open 12-lead ECG dataset. The experimental results show that MLBF-Net obtains an average $F_{1}$ score of 0.855, reaching the highest arrhythmia classification performance. The proposed method provides a promising solution for multi-lead ECG analysis from an information fusion perspective.

Objective: Myoelectric pattern recognition has been successfully applied as a human-machine interface to control robotic devices such as prostheses and exoskeletons, significantly improving the dexterity of myoelectric control. This study investigates the feasibility of applying myoelectric pattern recognition for controlling a robotic hand in stroke patients. Methods: Myoelectric pattern recognition of six hand motion patterns was performed using forearm electromyogram (EMG) signals in paretic side of 8 stroke subjects. Both the random cross validation (RCV) and the chronological handout validation (CHV) were applied to assess the offline myoelectric pattern recognition performance. Experiments on real-time myoelectric pattern recognition control of an exoskeleton robotic hand were also performed. Results: An average classification accuracy of 84.1% (the mean value from two different classifiers) and individual subject differences were observed in the offline myoelectric pattern recognition analysis using the RCV, while the accuracy decreased to 65.7% when the CHV was used. The stroke subjects achieved an average accuracy of 61.3-20.9% for controlling the robotic hand. However, our study did not reveal a clear correlation between the real-time control accuracy and the offline myoelectric pattern recognition performance, or any specific characteristics of the stroke subjects. Conclusion: The findings suggest that it is feasible to apply myoelectric pattern recognition to control the robotic hand in some but not all of the stroke patients. Each stroke subject should be individually online tested for the feasibility of applying myoelectric pattern recognition control for robot-assisted rehabilitation.

Objective: the objective of this work is to develop a robust method for myoelectric control towards alleviating the interference of electrode shift. Methods: In the proposed method, a preprocessing approach was first performed to convert high-density surface electromyogram (HD-sEMG) signals into a series of images, and the electrode shift appeared as pixel shift in these images. Next, a data augmentation approach was applied to the training data from just one position (no shift), so as to simulate HD-sEMG images derived from fictitious shift positions. The dilated convolutional neural network (DCNN) was subsequently adopted for classification. Compared to common convolutional neural network, DCNN always contained a larger receptive field that was supposed to be adept at mining wider spatial contextual information in images. This property was further confirmed to facilitate the classification of myoelectric patterns using HD-sEMG. The performance of the proposed method was evaluated with HD-sEMG data recorded by a 10 × 10 electrode array placed over forearm extensors of ten subjects during their performance of six wrist and finger extension tasks. Results: Under a variety of actual electrode shift conditions, the proposed method achieved a mean classification accuracy of 95.34%, and it outperformed other common methods. Conclusion: This work demonstrated feasibility and usability of combining data augmentation and DCNN in predicting myoelectric patterns in the context of electrode shifts. Significance: The proposed method is a practical solution for robust myoelectric control against electrode array shifts.

This study presents automatic decomposition of high density surface electromyogram (EMG) signals through a progressive FastICA peel-off (PFP) framework. By incorporating FastICA, constrained FastICA and a peel-off strategy, the PFP can progressively expand the set of motor unit spike trains contributing to the EMG signal. A series of signal processing techniques were applied and integrated in this study to automatically implement the two tasks that often require human operator interaction during application of the PFP framework, including extraction of motor unit spike trains from FastICA outputs and reliability judgement of the extracted motor units. Based on these advances, an automatic PFP (APFP) framework was consequently developed. The decomposition performance of APFP was validated using simulated high density surface EMG signals. The APFP was also evaluated with experimental surface EMG signals, and the decomposition results were comparable to those achieved from the PFP with human operator interaction.