SUPPORT VECTOR MACHINE CLASSIFICATION OF MUSCLE INTENSITY PATTERNS DURING PROLONGED RUNNING

INTRODUCTION Prolonged running results in altered muscle properties and control strategies that are expressed as changes in electromyographic (EMG) activity. Both central control and peripheral factors may change during the course of a fatiguing run and influence the frequency and amplitude of the EMG signal, respectively. In addition, EMG signals show precicely timed muscular events during a movement [1]. Our purpose was to test the hypothesis that the centrally and periferally controled features of the EMG signal change in a systematic way during a fatiguing run. If so, one should be able to discriminate between EMG measured in the non-fatigued or fatigued state. To discriminate between the early and late phases of endurance running we combined the wavelet based time-frequency analysis [2] with support vector machine classification. This study demontrates the ability of this analysis approach to quantify and classify data based on very subtle changes in the control of muscles.