Analyzing Speech Motor Movement using Surface Electromyography in Minimally Verbal Adults with Autism Spectrum Disorder

TL;DR

This study addresses the lack of direct physiological markers for speech motor control in minimally verbal adults with autism (mvASD) by applying eight-channel surface electromyography (sEMG) to facial muscles during diadochokinesis (DDK) and six-word tasks. It combines RMS power, inter-channel correlations, and time-delay embedded correlation matrices (size $120\times120$) to quantify neuromuscular activity and coordination complexity, reporting $N_{mvASD}=12$ vs. $N_{NT}=14$ with group differences including a significant increase in inter-channel correlation ($p=\$0.012$) and reduced complexity in mvASD as shown by eigenvalue spectra. Although RMS power differences were not consistently significant, mvASD displayed higher sEMG power and greater synchrony, suggesting tightly coupled, lower-DOF facial motor patterns during speech. These findings provide objective physiological markers of speech motor control in mvASD and support multimodal approaches for assessment and intervention, with future work extending to phoneme-specific analyses and integration with audio, video, and handwriting data ($N=12$, $14$).

Abstract

Adults who are minimally verbal with autism spectrum disorder (mvASD) have pronounced speech difficulties linked to impaired motor skills. Existing research and clinical assessments primarily use indirect methods such as standardized tests, video-based facial features, and handwriting tasks, which may not directly target speech-related motor skills. In this study, we measure activity from eight facial muscles associated with speech using surface electromyography (sEMG), during carefully designed tasks. The findings reveal a higher power in the sEMG signals and a significantly greater correlation between the sEMG channels in mvASD adults (N=12) compared to age and gender-matched neurotypical controls (N=14). This suggests stronger muscle activation and greater synchrony in the discharge patterns of motor units. Further, eigenvalues derived from correlation matrices indicate lower complexity in muscle coordination in mvASD, implying fewer degrees of freedom in motor control.

Figures (4)

• Figure 1: Simultaneous audio and sEMG data collected from an mvASD participant during speech production tasks revealing approximate synchrony across many channels. The different sEMG channels are color-coded.
• Figure 2: Placement of eight electrodes for surface electromyography (sEMG) with two reference electrodes on the ears.
• Figure 3: Correlation between the sEMG channels
• Figure 4: Cohen D effect sizes for eigenvalues derived from auto- and cross-correlation matrices during speech tasks.