Identification and Classification of Myoelectric Signal Features Related to Hand Motions

Either a hand movement or a gesture appears to be worthy of study regarding industrial requirements for operators who have to accomplish multiple activities with a high recurrence. We propose a pattern recognition system for the categorization of hand motions based on the technique of linear discriminant analysis (LDA) of EMG phenomena. Because LDA is a statistical approach allowing for simultaneous assessment of the differences between two or more groups regarding many variables or sets of variables, it is being used for accurate evaluation of the muscle-force relationship. In this investigation, we used surface electromyogram (sEMG) data collected from ten volunteers. sEMGs were recorded via two muscle channels (m. flexor digitorum and m. extensor digitorum). Matlab (R) was used to extract features and other necessary parameters, and further statistical analysis in the form of pairwise comparisons was performed using SPSS (R). An efficiency of 88.6 and 87.1% was provided by the proposed system regarding channel 1 and channel 2 muscle locations respectively. Further, these results may have an essential value for researchers actively involved in hand prosthetic design.