Subject-Independent sEMG Pattern Recognition by Using a Muscle Source Activation Model

Abstract

The interpretation of surface electromyographic (sEMG) signals facilitates intuitive gesture recognition. However, sEMG signals are highly dependent on measurement conditions. The relationship between sEMG signals and gestures identified from a specific subject cannot be applied to other subjects owing to anatomical differences between the subjects. Furthermore, an sEMG signal varies even according to the electrode placement on the same subject. These limitations reduce the practicability of sEMG signal applications. This letter proposes a subject-independent gesture recognition method based on a muscle source activation model; a reference source model facilitates parameter transfer from a specific subject, i.e., donor to any subject, donee. The proposed method can compensate for the angular difference of the interface between subjects. A donee only needs to perform ulnar deviation for approximately 2s for the overall process. Ten subjects participated in the experiment, and the results show that, in the best configuration, the subject-independent classifier achieved a reasonable accuracy of 78.3% compared with the subject-specific classifier (88.7%) for four wrist/hand motions.