Surface Electromyography Based Finger Flexion Recognition

Great effort is being taken since the last decade to make control interfaces for machines and robots less complicated and more natural, by taking signals directly from the body. This requires highly accurate signal interpretation. Researches have shown that electromyographic signals obtained from skeletal muscles can give very accurate information regarding limb movements, that can be used in human robot interface. In this paper, a novel approach toward classifying finger movements using surface Electromyography (sEMG) is discussed. Higuchi fractal dimensions and auto-regressive modeling are employed for extracting suitable features from a 6-channel EMG signal acquired indigenously using self-designed data acquisition hardware and classified using Multi Class Support Vector Machine. An average accuracy of 95.2% has been obtained.