Body Motion Noise Reduction of Silent Speech Recognition Using Facial Surface EMG

This paper proposes a method for reducing body motion noise in silent speech recognition (SSR) systems using facial muscle information. Several SSR methods have been developed that utilize facial muscle information for speech recognition. However, these methods are limited to steady-state conditions. The accuracy of these methods is significantly degraded by body motion noise from daily movements mixed with EMG signals, making word identification impossible under such conditions. To address the body motion problem in SSR systems, this paper proposes an improved SSR system and demonstrates its effectiveness.