A pattern recognition research for crosswise normalized forearm SEMG signal

Abstract

SEMG (surface electromyogram) signal is the electrical activity of human body movement, different SEMG is the characterization of the different movements. This paper analyzes the collected SEMG by time-domain method, extracted time domain characteristic value, constructed the characteristic value vector of multiple parameters before and after normalization, using the average value as the training sample, and then makes the pattern recognition to the SEMG of the forearm and hand four different actions based on BP neural network. The results show that the normalized time-domain has a better recognition effect, and this could have certain practical reference value for the SEMG controlled artificial limb.