Modeling and classification of sEMG based on instrumental variable identification

Abstract

sEMG is biological signal produced by muscular. According to the characteristics of myoelectric signal, FIR model of the single input and multiple outputs was proposed in this paper. Due to the unknown input of the model, instrumental variable with blind identification was used to identify the model's transfer function. The parameters of model were used as input of neural network to classify six types of forearm motions: extension of thumb, extension of wrist, flexion of wrist, fist grasp, side flexion of wrist, extension of palm. The experimental results demonstrate that this method has better classification accuracy than the classical AR method. © 2011 Springer-Verlag.