Relative Confidence Based Information Fusion For Semg-Based Pattern Recognition

Abstract

This study aims to establish a decision fusion scheme that synthesize the advantages of different classifiers and avoids uncertain decisions. Thus, relative confidence factors of each classifier was proposed to correct the classification decision made by each classifier, and the final classification result was derived by fusing all corrected decision based on the combination of Dempster-Shafer's rule. The novel fusion scheme is evaluated in the scenario of sEMG-based hand movement identification, in which five classffiers are adopted. The experimental results demonstrated that the novel scheme can obtain higher classification accuracy and stability than the other methods.