Skip to Main Content
The classification of electromiographic signals provides a method to identify one's intent or movement. This paper presents a study on choosing the best model for a classifier used to group electromiographic signals into classes corresponding to the isometric flexion effort of different fingers. The signals were collected from the Flexor Digitorum Superficialis and Profundus of seven healthy subjects. Different features (root mean square -RMS, average rectified value - ARV, mean and median frequency) and different classifier structures (discriminant analysis, nearest neighbor analysis, naive Bayes algorithm, neural network, fuzzy logic based algorithm) were implemented with classification success rates ranging from 50 to 99 %. The success rate of the classifiers corresponds to the ability of a numerical system to decode the physiological manifestations associated with the finger movements.