EMG-EMG correlation analysis for human hand movements

In this paper, a novel electromyogram (EMG)-EMG correlation analysis method is proposed to identify human hand movements. Mutual information (MI) measure is employed to analyse the ordinal pattern of the surface EMG recordings. The MI measure is extracted from EMG signals and compared with other various sEMG features in the time and frequency domains. The comparative experimental results demonstrate that autoregressive coefficients (AR)+MI has a better performance than the single features and other multi-features. The multi-features combining the different features mostly have improved the recognition performance, and the MI provides important supplemental information to the hand movements. It is evident that the proposed correlation feature is essential to improve the recognition rate.