Temporal vs. spectral approach to feature extraction from prehensile EMG signals

Abstract

There are generally two nonparametric approaches in feature extraction from temporal signals: temporal and spectral approach. Both approaches were used in classification of prehensile electromyographic (EMG) signals. The goal of this paper is to define and evaluate some successful methods in both approaches and to determine experimentally which method and approach is the most appropriate. The evaluation is based on classification of real EMGs with an ART-based classifier. The efficiency analysis is also provided. The results have shown that a less expensive temporal approach has strong advantages over the spectral methods.