Characterizing Speech Motor Pattern in Minimally Verbal Adults with Autism Spectrum Disorder via Surface Electromyography

Abstract

Minimally verbal adults with Autism Spectrum Disorder (mvASD) experience significant speech production challenges linked to impaired motor skills. Despite the prevalence of these speech difficulties, the underlying motor mechanisms remain poorly understood. This thesis investigates the neuromuscular activity associated with speech motor movement in mvASD using surface electromyography (sEMG). By capturing and analyzing sEMG signals with 8 electrodes from key facial muscles during speech production tasks, this study provides insights into the distinct motor patterns exhibited by mvASD individuals compared to neurotypical controls. The sEMG data was collected while 25 participants, including 10 mvASD individuals and 15 neurotypical controls performed a series of carefully designed speech tasks. Features such as Root Mean Square (RMS) values, Pearson correlation coefficients, and eigenvalues from auto and cross correlation matrices were extracted to measure muscle activation and coordination complexity. The results reveal that mvASD individuals exhibit higher RMS values and greater synchronization between sEMG channels, indicating stronger muscle activation and tighter coupling among facial muscles. Furthermore, the analysis of eigenvalues suggests lower complexity in motor coordination among mvASD participants, reflecting fewer degrees of freedom in muscle control. These findings were supported by classification models, which demonstrated that features from diadochokinetic tasks were more effective in distinguishing mvASD from neurotypical individuals.