A hybrid AB-RBF classifier for surface electromyography classification

Abstract

In this paper, we aim to classify surface electromyography (sEMG) by using Attribute Bagging-Radial Basis Function (AB-RBF) hybrid classifier. Eight normally-limbed individuals were recruited to participate in the experiments. Each subject was instructed to perform six kinds of finger movements and each movement was repeated 50 times. Features were extracted using wavelet transform and used to train the RBF classifier and the AB-RBF hybrid classifier. The experiment results showed that AB-RBF hybrid classifier achieved higher discrimination accuracy and stability than single RBF classifier. It proves that integrating classifiers using random feature subsets is an effective method to improve the performance of the pattern recognition system. © Springer-Verlag Berlin Heidelberg 2007.