Analysis of dynamic change in sampling rate of EMG signal for designing prosthesis control

Electromyograph (EMG)-based prosthetic devices use a classifier to identify the myoelectric signals obtained from the amputee's limb. The classifier's output then controls the movement of the prosthetic device. The spectrum of the EMG signal is known to dynamic change with muscle length. These changes are reflected in the classifier's output, and so they are an important component in constructing a strong classification system to determine the user's intent. This work presents a method for dynamically changing the sample frequency of the EMG based on the spectrum of the input myoelectric signal. The shift in sampling frequency allows for correction of muscle length variations and the resulting resilience in signal classification. The approach was successfully validated and implemented on simulated EMG. The results indicate that variations in classifier output caused by changes in muscle length may be compensated for by changes in the spectrum.