Multiclassifiers with competence function applied to the recognition of EMG signals for the control of bio-prosthetic hand

The paper presents a concept of bio-prosthesis control via recognition of user intent on the basis of myopotentials acquired of his body. We assume that in the control process each prosthesis operation consists of specific sequence of elementary actions. The multiclassifier systems with fusion/selection strategy based on competence function are applied to the recognition of patient's intent. Experimental investigations of the proposed multiclassifiers for real data are performed and results are discussed. Classification results obtained for three simple fusion methods and one multiclassifier system are used for a comparison. ©2009 IEEE.