Classifier fusion interactive software toolbox for EMG signal decomposition

Abstract

A classifier fusion interactive software package has been constructed for implementing the classification task in the electromyographic (EMG) signal decomposition process using the MATLAB high-level programming language and its interactive environment. The package employs classifier fusion schemes of multiple classifier combination for the purpose of fusing the decisions of a set of heterogeneous base classifiers to make a final decision that achieves improved classification performance. The base classifiers used are ensembles of error-independent certainty, fuzzy k-NN, and template matched filter classifiers. The interactive package consists of several graphical user interfaces (GUIs) to extract individual motor unit potential (MUP) waveforms from raw EMG signals; extract relevant features; classify MUPs into motor unit potential trains (MUPTs) using certainty-based, assertion-based, and similarity-based classifiers; and combine classifier decisions. The proposed software package is useful for enhancing the EMG signal analysis quality and providing a systematic approach to the EMG signal decomposition process. It worked as a very helpful environment for testing and evaluating algorithms developed for EMG signal decomposition research.