Arm movement recognition from EMG signals using HHT based method

The aim of this study is to propose a procedure for the recognition of arm movements using EMG signals. This recognition will be used to control intelligent arm prostheses by the brain via EMG signals collected near the absent limb.Real EMG records were used. The proposed method consists in extracting relevant intrinsic mode functions (IMF), calculating their HHT, and then extracting the most relevant features. This method provided short length vectors of relevant features (11) able to distinguish between 6 pairs of 4 arm movements: flexion, extension, pronation and supination. The obtained mean accuracy values from linear SVM classification for 5/6 motion pairs were between 93,11% and 97.95% and average AUC values exceeded 98.06%.These results are satisfying and the efficiency of our method is thereby proved in even the recognition of movements domain.