Real-Time Chinese Sign Language Recognition Based on Artificial Neural Networks
Zhen Zhang, Ziyi Su, Ge Yang · 2019
Sign language is the main method for the deaf-mute to communicate with each other. However, normal people cannot understand the meaning of specific gestures without training. In this study, a real-time Chinese Sign Language (CSL) recognition model based on surface electromyographic (sEMG) signals and Artificial Neural Networks (ANN) is proposed. The model achieves an average accuracy at 88.7% on 15 CSL gestures with average response time around 300ms (timing at movement begins). In our approach, the MYO armband is used for acquiring raw sEMG signals. Signal preprocessing includes rectification, filter and extracting the muscle activation section. We apply a sliding window and a muscle detection approach to segment the signals and extract features. A test method is used to recognize the gesture when the totality of same label, which returned by ANN classifier, reach the activation times threshold.