Research on Chinese Sign Language Recognition Methods Based on Mechanomyogram Signals Analysis

This paper presents an integrated approach to Chinese sign language (CSL) actions recognition, which involves Teager-Kaiser energy operator (TKEO) segmentation, wavelet feature extraction and support vector machine (SVM) classification on mechanomyogram (MMG) Signals. It used a four-channel wireless signal acquisition system to collect the MMG signals of the extensor digitorum (ED), flexor carpi radialis (FCR), flexor carpi ulnaris (ECU) and extensor carpi radialis (ECR). After filtering, the TKEO algorithm was used to segment the MMG signals. The wavelet packet energy (WPE) of MMG signals were extracted as features for further analysis. SVM was applied as a classifier to recognize 18 CSL actions. Compared with other commonly used methods, the proposed method had better recognition accuracy and recognition performance as well. The average recognition accuracy of the proposed method was up to 95.38%.