IDENTIFICATION OF EMG FREQUENCY PATTERNS IN RUNNING BY WAVELET ANALYSIS AND SUPPORT VECTOR MACHINES

Abstract

The purpose of this study was to identify EMG pattern of running at different speed and incline based on a trial-to-trial analysis. Eight subjects performed treadmill running at five different conditions (4, 5 and 6 m/s, 5m/s at 5° incline, 5m/s at 2° decline). EMG data of eight leg muscles were recorded and transformed by a wavelet analysis (van Tscharner, 2000). Ten subsequent steps of each subject and condition were classified by support vector machines. Between 93 and 100% of all EMG patterns were assigned correctly to the individual. According to the different running conditions recognition rates ranged between 78 and 88%. Hence, support vector machines can be considered as powerful nonlinear tool for the classification of dynamic EMG patterns.