Articulator Muscle Contraction Evaluation in Vowel Phonemes Articulation Based on Facial Images and Electromyography Signal

Utilization of motor signal information of articulatory muscles and facial images separately in speech recognition shows that these two pieces of information are related to the articulation process. This study examines the relationship between EMG signals and facial deformation in several conditions during the pronunciation of vowel phonemes. The characteristics of the articulatory muscle contractions around the oral cavity are reviewed for 3 phases (initial, maximum, and final) for the pronunciation of each vowel phoneme. Segmentation and wavelet denoising are used to condition the EMG signal; then, the RMS amplitude is extracted from the signal using an RMS envelope. 2D vector maps are extracted from 3-phase images and displayed as compass maps for each phase. In this study, the relationship between motor signals of articulation muscles and facial articulator movements was found. This finding is shown in all pronunciations from EMG and vector values, which are the largest in phase 2. Also, the activity of muscle depressor anguli oris plays more significant role in the pronunciation of vowel phonemes.