The joint application of rough set-based feature reduction and information fusion in motion recognition

An effective and efficient strategy for motion classification via the electromyographic (EMG) signals is proposed in this paper. The wavelet packet transform (WPT) is utilized to extract the energy characteristics of the sub-bands as the features. Given the redundancy between the features, the rough set theory (RST) is employed to implement the feature reduction. For improving the classification accuracy, the multi-classifier based information fusion strategy is performed to avoid the arbitrary decision of only one classifier with insufficiency information. Compared with the single-classifiers using the same features, more excellent performance indicates the potential of the RST-based feature reduction and the multi-classifier fusion techniques in motion classification.