Exploring Muscle Synergy Consistency and Physiological Interpretability in Speech Evaluation Using High-Density sEMG.

Abstract

Speech production involves intricate muscle coordination, presenting unique challenges for motor control analysis. This study systematically explores muscle synergies in speech production using high-density surface electromyography (HD-sEMG), providing novel insights into how muscle groups coordinate during speech tasks. Ten healthy participants performed 14 vowel and 15 consonant tasks in both phonated and silent speech modes. The muscle activity was recorded using five HD electrode arrays covering the face, jaw, and neck. The study identified three key muscle synergies sufficient to reconstruct muscle activation patterns with over 95% variance accounted for (VAF). These synergies demonstrated hierarchical and collaborative roles: the dominant synergy controlled core articulatory movements, the complementary synergy supported auxiliary tasks, and the modulatory synergy refined fine motor control. Synergies exhibited high intra-subject stability (cosine similarity $>$ 0.7) and moderate inter-subject reproducibility, reflecting shared coordination strategies alongside individual variability. Synergy-based features combined with traditional time-domain features improved classification performance, achieving accuracies of up to 94.72%, highlighting the complementary value of synergy-derived spatial information. These findings connect motor control theories with speech production research, advancing our understanding of the biomechanical basis of vocalization. The results lay a foundation for developing synergy-based human-machine interfaces for applications in speech prosthetics, recognition, and rehabilitation.