SEMG classification for upper-limb prosthesis control using higher order statistics

The aim of this paper is to present application of higher order statistics for surface electromyogram (sEMG) signal pattern classification. The new pattern recognition algorithm exploits a multilayer perceptron (MLP) as the classifier and the feature vector is a combination of cumulants of the second-, third- and fourth- orders and integral of absolute (IAV) of two channel sEMG stationary segments. The detected sEMG signals are used in classifying four upper-limb primitive motions, namely, elbow flexion (F), elbow extension (E), wrist supination (S) and wrist pronation (P). The simulation results illustrate the considerable accuracy of the proposed framework in sEMG pattern recognition.