Towards Chinese sign language recognition using surface electromyography and accelerometers

摘要

Sign language recognition (SLR) could help the deaf better communicate with individuals who do not understand sign language, which could also be used in human-computer interaction (HCI). In this paper, surface Electromyography (sEMG) signals and accelerometer (ACC) signals are acquired from the right forearm, wrist and the back of the hand for 18 isolated Chinese Sign Language (CSL) signs. A new position to collect sEMG and ACC, which is on the back of the hand, is proposed to improve recognition rate. Also, features that are related to the attitude are extracted to further improve the recognition rate. The improved method is evaluated on 8 healthy subjects. Experimental results showed that fusing sEMG and ACC on the back of the hand, and extracting the attitude related features could improve the recognition rate in a statistically significant way. This method could increase the recognition rate from 84.9% to 91.4% in a window length of 176 ms.