Multiclassifier System with Fuzzy Inference Method Applied to the Recognition of Biosignals in the Control of Bioprosthetic Hand

摘要

The paper presents an original method of recognition of patient's intention to move of hand prosthesis during the grasping and manipulation of objects. The proposed method is based on a 2-level multiclassier system (MCS) with base classifiers dedicated to EMG and MMG signals, and with combining mechanism using a dynamic ensemble selection (DES) scheme and competence function. Competence function of base classifier is determined using validation set in the two step procedure. The first step consists in creating competence set using the methods based on relating the response of the classifier with the response obtained by a random guessing. In the second step, the competence set is generalized to the whole feature space using the learning procedure based on the Mamdani-type fuzzy inference system. The performance of MCS with proposed competence measure was experimentally compared against four benchmark classification methods using real data concerning the recognition of six types of grasping movements. The system developed achieved the highest classification accuracies demonstrating the potential of MC system for the control of bioprosthetic hand.