Skip to main content

27-03-2013 | Psychology | Article

Automated speech analysis estimates vocal development in ASD


Free abstract

medwireNews: US study findings suggest that day-long recordings of vocalizations from children with autism spectrum disorder (ASD) can provide a reliable estimate of vocal development that is related to expressive spoken language ability.

Paul Yoder (Vanderbilt University, Nashville, Tennessee) and colleagues say the research validates a recently developed vocal measure that is derived from automatic analysis of sample voice recordings.

According to the researchers, automated measures have the potential to quantify a greater number of speech-like vocalization aspects compared with traditional measures that require human perception.

"If future research continues to support the construct validity of this new measure, we may have greater ability to predict which children with ASD will learn to talk... and have improved ability to predict whether a child is ready for spoken language intervention," write the authors in Autism Research.

In total, 40 children with ASD and 29 healthy controls wore a small digital language processor for 3 days within a 10-day period. Vocal samples were then analyzed using existing software, which assesses vocalizations against 12 parameters of speech.

These parameters quantify various acoustic products of rhythmic movements of articulators that produce syllabic sounds, voice quality characteristics found in speech, and pitch and duration characteristics consistent with speech.

Study findings showed that vocal development age measurements were highly stable when assessing recordings from 1 day in both ASD and typically developing children.

Given that most behaviorally measured variables in young children with ASD are unstable, the current results are significant say the researchers.

Vocal development measures in both ASD and control children were significantly associated with the number of words parents reported their children said on the Language Development Survey and with the age equivalency score derived from the Child Development Inventory, Expressive Language Scale.

According to the authors, the high correlations seen with expressive language indicate that the automatically generated age equivalency score is relevant to, but not a measure of, expressive language.

"Aside from the advantage of objectivity, the lack of necessity to involve human perception in generating the vocal development age equivalency score reduces threat to the privacy of participants, and provides great potential for access for non-technical users," conclude Yoder and team.

By Ingrid Grasmo, medwireNews Reporter

Related topics