Speech Characteristics of Young Children with ASD

Background: Autism spectrum disorder (ASD) is a neurodevelopmental disorder which is characterized by deficits in social interaction and the presence of repetitive behaviors and restricted interests. Previous studies have reported that children with ASD exhibit a variety of speech abnormalities in comparison to typically developing children. Here we analyzed recordings of speech from ADOS assessments using signal processing techniques. We quantified several prosodic and structure characteristics in individual children and examined their relationship with ASD severity.

Methods: Audio recordings of 72 children, 2-7-years-old, were extracted from ADOS assessments that were performed at the National Autism Research Center of Israel. The recordings were manually annotated to identify segments of the child’s speech and the clinician’s speech during the ADOS assessment. Vocal islands of the child’s speech were then identified by their energy and a variety of prosodic features were calculated within each vocal island.

Results: Significant negative correlations were found between the SA scores of the ADOS assessment and measures of speech structure, such as the number of responses (r=-0.45, p<0.001). Significant positive correlations were found between the RRB scores of the ADOS and pitch features (i.e. mean pitch (r=0.58, p<0.001)).

Conclusions: These results demonstrate that recordings of the child’s interaction with the clinician during the initial ADOS assessment hold important information regarding the children’s ASD severity that can be extracted using speech processing techniques. This highlights the potential clinical utility of speech analysis for estimating ASD symptom severity at a very young age.