A variety of electromyographic (EMG) features have been evaluated for
control of myoelectric cybernetic prosthetic arms. Movement class disc
rimination, robustness, and computational complexity of these features
have been investigated for different window sizes and noise levels. A
n adaptive selection of window size for feature extraction has also be
en developed and evaluated. The experiments were done on the data acqu
ired from the residual biceps and triceps muscle of an above-elbow amp
utee.