Self-powered speech recognition system for deaf users

Abstract

Some members of the deaf community suffer inconveniences in daily life due to communication difficulties. In many cases, deaf people’s vocal cords are intact, and so establishing speech recognition based on the correspondence between their intention and vocal cord vibration may help overcome some communication barriers, in particular human-machine interactions. Here, we report an anti-interference speech recognition system intended primarily for deaf people using a self-powered triboelectric vibration sensor (STVS) assisted by a contextual recognition model (CRM). A soft and safe woven structure nanofiber cellulose film (NFCF) is prepared as a vibration-sensitive layer, which endows STVS with high sensitivity at wide vibration frequencies (up to 1,000 Hz). The CRM can identify 17 commonly used expressions with an accuracy above 92.3% and provide voice identity recognition with 90.6% accuracy. Our speech recognition system could help in providing a facile, convenient, and bidirectional communication channel between deaf people and hearing people.