Sign Language Vocabulary Recognition Only with Tactile Sensing Glove
Motoki Kagami, Zeping Yu, Sim Teck Ceng, Lei Jing · 2024
A significant challenge in communication arises between deaf individuals and hearing individuals. Sign language recognition is expected to be an important research field for eliminating barriers. In this field, data collection is primarily conducted using two methods: sensor-based and vision-based approaches. In this study, we adopted a sensor-based approach by using a tactile sensing glove due to its portability, costeffectiveness, and ability to capture fine finger movements. Subsequently, we applied LSTM and k-NN respectively to evaluate the recognition accuracy using this method. As a result, LSTM achieved a test accuracy of approximately 76%, while k-NN demonstrated an average accuracy of 87%. These findings highlight the potential of sensor-based SLR technology in reducing communication barriers and advancing inclusively in communication tools.