Development of low cost EMG data acquisition system for Arm Activities Recognition

Abstract

Electromyography (EMG) signals are becoming continuously more important in many fields, including biomedical/clinical, prosthesis, human machine interaction and rehabilitation devices. Due to its undesirability to drive display and monitoring systems, it needs an adaptable and robust enough processing units to clearly picture the signal. In the present study, to meet the requisites of EMG data acquisition systems, a high resolution, and highly competitive eight channel system has been developed, which is cost efficient and compact as compared to commercially available systems. To validate the developed system, EMG signals have been acquired from various muscles for different arm activities and also machine learning techniques were utilized for activity recognition. For the current study 8 Male and 4 Female healthy subjects were recruited. For classification purpose various time and frequency domain features were extracted and two kind of feature selection techniques were utilized. The classification accuracy ranges from 43.64% to 92.61% for different classification algorithms. Further comparative study of different classification techniques is presented. After the selection of features accuracy 92.66% has been achieved with only two muscles’s contribution and this work has been compared with previous research. For this piece of work MATLAB 15a is utilized for signal processing and machine learning.