Prediction of sEMG-based tremor joint angle using the RBF neural network

Pathological tremor includes many types and affects person´s normal life. To suppress the tremor, the medical drugs or auxiliary exoskeleton system were applied. The amplitude and frequency of tremor movement were crucial to effective suppression by exoskeleton system. In this study, base on the drawback of the kinematic information and the relationship between surface electromyographic (sEMG) signals and muscle activations, the prediction of angle from sEMG signals by using network was proposed and realized, avoiding establishing a complex mathematical model to describe the angular displacement. Comparing the sEMG captured from tremor with the normal signals, the linear profile-curve of sEMG was extracted as a feature value, and the relationship model between joint angle and sEMG was constructed by the radial basis function neural network (RBF). The analysis of the experimental data showed that the linear profile-curve could be used to predict the angle or angular velocity, with higher accuracy and adaptability comparing with root mean square (RMS) of EMG.