Search

Traditional myoelectric prostheses that employ a static pattern recognition model to identify human movement intention from surface electromyography (sEMG) signals hardly adapt to the changes in the sEMG characteristics caused by interferences from daily activities, which hinders the clinical applications of such prostheses. In this paper, we focus on methods to reduce or eliminate the impacts of three types of daily interferences on myoelectric pattern recognition (MPR), i.e., outlier motion, muscle fatigue, and electrode doffing/donning. We constructed an adaptive incremental hybrid classifier (AIHC) by combining one-class support vector data description and multi-class linear discriminant analysis in conjunction with two specific update schemes. We developed an AIHC-based MPR strategy to improve the robustness of MPR against the three interferences. Extensive experiments on hand-motion recognition were conducted to demonstrate the performance of the proposed method. Experimental results show that the AIHC has significant advantages over non-adaptive classifiers under various interferences, with improvements in the classification accuracy ranging from 7.1% to 39% ( ${p} ). The additional evaluations on data deviations demonstrate that the AIHC can accommodate large-scale changes in the sEMG characteristics, revealing the potential of the AIHC-based MPR strategy in the development of clinical myoelectric prostheses.

In video-based action recognition, using videos with different frame numbers to train a two-stream network can result in data skew problems. Moreover, extracting the key frames from a video is crucial for improving the training and recognition efficiency of action recognition systems. However, previous works suffer from problems of information loss and optical-flow interference when handling videos with different frame numbers. In this paper, an augmented two-stream network (ATSNet) is proposed to achieve robust action recognition. A frame-number-unified strategy is first incorporated into the temporal stream network to unify the frame numbers of videos. Subsequently, the grayscale statistics of the optical-flow images are extracted to filter out any invalid optical-flow images and produce the dynamic fusion weights for the two branch networks to adapt to different action videos. Experiments conducted on the UCF101 dataset demonstrate that ATSNet outperforms previously defined methods, improving the recognition accuracy by 1.13%.

Decomposition of EMG (electromyography), in which the MUAPs (Motor Units Action Potentials) are extracted from the original EMG signals, is a useful approach for evaluating the physiological properties of muscles and studying the neural mechanism of human motions. Existing decomposition methods usually extracted the MUAPs via the waveform matching between the detected spikes and MUAP templates. However, those methods always involve too many artificial parameters and have no generally mathematical expression, which make them inflexible in practical applications and also cause poor robustness under interference. With respect to the problem, in this paper, a wavelet-based decomposition method is proposed to detect the “true” MUAPs from sEMG. Two standard Gaussian wavelets are firstly defined as the basic MUAP expressions; then, wavelet transform is performed to describe scales and magnitudes of different MUAPs from different nerve fibers. Experiments were conducted to verify the performance of the proposed method, where MUAPs decomposed from sEMG under different levels of muscle contraction can be effectively detected by the method.

Traditional EMG-motion recognition methods are only able to recognize target motions that presented in the training phase, but cannot detect randomly-occurred outlier motions that did not present before. Here, a hybrid classifier that combines one-class SVMs and a multi-class LDA was proposed to perform recognition on target classes and rejection on outlier classes. The classification ability of the hybrid classifier can incrementally grow via online learning the data of outlier classes. Extensive experiments on EMG-based hand-motion recognition were conducted to verify the performance of the incremental hybrid classifier (IHC). The mean recognition accuracy on target classes of IHC is 92%, which is 23% higher than that of the normal MLP. Moreover, IHC has the ability to detect outlier patterns that MLP would misclassify to target classes.

Missing data are a common drawback that pattern recognition techniques need to handle when solving real-life classification tasks. This paper first discusses problems in handling high-dimensional samples with missing values by the Gaussian mixture model (GMM). Since fitting the GMM by directly using high-dimensional samples as inputs is difficult due to the convergence and stability issues, a novel method is proposed to build the high-dimensional GMM by extending a reduced-dimensional GMM to the full-dimensional space. Based on the extended full-dimensional GMM, two approaches, namely, marginalization and conditional-mean imputation, are proposed to classify samples with missing data in online phase. Then, the proposed methods were employed to recognize hand motions from surface electromyography (sEMG) signals, and more than 75% of classification accuracy of motions can be obtained even if 50% of sEMG signals were missing. Comparisons with normal mean and zero imputations also demonstrate the improvements of the proposed methods. Finally, a control scheme for a myoelectric hand was designed by involving the novel methods, and online experiments confirm the ability of the proposed methods to improve the safety and stability of practical systems.

Electromyography (EMG) has been popularly used as interface command to achieve a natural control for myoelectric prosthetic-hands. Traditional EMG-based recognition methods always only focus on the classification of target motion classes that were defined in the training phase, but have no ability to reject outlier motion interferences that did not present before. In this paper, a hybrid classifier that combines one one-class Gaussian classifiers and a multi-class LDA was constructed to achieve EMG-based motion classification, in which Gaussian classifiers were used to reject outlier interferences, while LDA was used to classify target motion samples. The robust hybrid classifier is easily built and has low run-time complexity. Extensive experiments were conducted to verify the performance of the proposed hybrid classifier, where 91.6% of target motion recognition accuracy and 96.5% of outlier motion rejection accuracy were respectively obtained. Finally, the hybrid classifier was involved to achieve a robust and real-time control of a myoelectric prosthetic-hand.

sEMG recognition has been used extensively in prosthetic device control, human-assisting manipulators and sign language recognition, etc. However, the sEMG recognition model, trained with one subject's sEMG data, is not applicable to the other subjects, which hinders the practical application of myoelectric interfaces immensely. In this paper, a sEMG recognition method which is applicable to multi-users is proposed. Firstly, single channel sEMG is decomposed into 30 MUAPTs, which includes four steps: two-order differential filter, threshold calculation, spike detection and hierarchical clustering. Secondly, the MUAPTs are updated with the templates orthogonalization; and Deep Boltzman Machine is employed to classify the MUAPTs into five classes corresponding to the predefined five gestures. Six participants participated in this experiment to validate the effectiveness of the proposed method. Results indicated that this method can achieve a mean accuracy of 81.5%.

Predicting Torque by using Electromyography (EMG) signals is a significant problem as long as wearable robots is concerned. In this paper, a coarse-to-fine algorithm, named as Hierarchical Projected Regression, is proposed to estimate the torque of elbow by the surface EMG signals from human's arm. HPR is learned based on a clinical dataset. First, the real torque distribution can be clustered by k-means algorithm. Meanwhile, the EMG signals are divided into multi-clusters corresponding to the partitions of torque dataset. Then, a projected matrix is derived based on the criterion of Linear Discriminant Analysis, by which the low dimensional features are extracted from the original high dimensional EMG. Finally, the process extends to multi-level tree algorithm based on the hierarchical variance reduction of individual torque cluster. The experimental results show the high accuracy and efficiency of this novel method.