| Title |
EMG-based Active Prosthetic Hand with Improved Convenience and Security through Speech Recognition and Speaker Verification |
| Authors |
김기승(Ki-Seung Kim) ; 김선홍(Seon-Hong Kim) ; 조용운(Yong-Un Jo) ; 오도창(Do-Chang Oh) |
| DOI |
https://doi.org/10.5370/KIEE.2022.71.2.443 |
| Keywords |
Speech recognition; Speaker verification; Convolutional Neural Network; Mel-Frequency Cepstral Coefficient; EMG; Active prosthetic hand |
| Abstract |
Electronic prosthetics using the existing electromyography (EMG) have the inconvenience of using external elements such as button manipulation to implement various hand movements, and there is a problem in that it is impossible to control the prosthetic hand using the EMG in a situation in which malfunctions may occur due to causes such as sensor failure or misalignment of the wearing position. In this study, convenience and security are improved by combining speech recognition and speaker verification technology with an prosthetic hand using EMG. Speech recognition was implemented as a function to convert speech into text (STT) through the speech recognition module of Python, and deep learning of the Convolutional Neural Network (CNN) structure was used for speaker verification. In the experiment, the Mel-Frequency Cepstral Coefficient (MFCC) was applied to the voice data for the four hand gestures collected from a total of 10 subjects to extract features, and then, learning and testing were conducted using these data as input to the CNN. As a result of the experiment, an average of 97% speaker verification accuracy was obtained for the four hand gestures. |