Hand motion identification from surface EMG signals based on multi-feature integration

Surface electromyography (sEMG) signals have been used as the inputs of many human-machine interface applications. Different methods for upper limb motion recognition are developed based on sEMG features. The sEMG feature extraction is the key issue in affecting classification performance in multi-motion pattern recognition. A number of features have been extracted to form feature combinations. However, these feature combinations are simple concatenation of multiple features. The combination weights for different sEMG features are equal. It may not help to design an effective multi-motion recognition system. In this study, to overcome the limitation and combine flexibly multiple features, we propose a multi-feature integration method based on multiple kernel learning (MKL) framework for motion recognition. The proposed method integrates the valuable information which can represent the motion patterns. Multiple kinds of features are fused together with optimal combination weights. In the experiment, we test the proposed method on multiple different hand motions. The corresponding experimental results show that the proposed method provides better classification performance than the other frequently used methods of feature combination. The results confirm that the proposed method is quite promising in integrating sEMG multi-features to advance the classification performance further for multi-motion recognition.