Muscle Fatigue Identification in Temporal -Spatial-Spectral Joint Domain

In order to effectively analyze the multi-dimensional feature of surface electromyography (sEMG) signals under fatigue state, a fatigue identification method based on temporal-spatial-spectral joint domain is proposed in this paper. Firstly, the sEMG signals of 10 healthy male subjects during grip strength are collected. Secondly, the complex Morlet wavelet transform is used to construct the multi-order tensor data with time, space, frequency and task information, and then the tensor linear Laplace discriminant analysis method is employed to obtain the best projection matrix. The training set and test set are projected by the projection matrix respectively to obtain the features with large discrimination. Finally, linear discriminant analysis (LDA) is adopted to classify the grasping action under normal and fatigue conditions. Experimental results show that the proposed method can achieve better average classification accuracy, compared with the root mean square, autoregressive coefficient and tensor high-order discriminant analysis methods. It is thereby shown that our method can effectively detect different states of muscle contraction.