Chief Science Officer / Fellow
Piezoelectric sensing of mask surface waves for noise-suppressive speech input
Author
Hiraki, Hirotaka and Rekimoto, Jun
Abstract
Voice-controlled medical devices in operating rooms face challenges due to face masks and noise, hindering speech recognition accuracy. We present a novel solution: a detachable piezoelectric microphone attached to a face mask, prioritizing hygiene and comfort. By capturing surface vibrations from the user’s voice, our system significantly improves scale-invariant signal-to-distortion ratio (Si-SNR) compared to traditional unidirectional microphones, especially for whispered speech, as demonstrated in a simulated noisy environment. This real-time, noise-suppressive approach offers a promising avenue for enhancing voice input and conversations in various applications and settings.