Acoustic changes in speech prosody produced by children with autism after robot-assisted speech training

Abstract

Children with Autism exhibit distinct speech prosody, perceived as monotone and they are reported to show deficits in focus marking. The current study designed a robot-assisted training with controlled social interactions aiming to enhance the prosody of children with Autism speaking a tonal language, Cantonese, specifically on focus marking. 20 autistic and 23 typically-developing (TD) children participated in this study. Only the autistic group received training. Stimuli were designed for training, pre- and post-training production. Acoustics of target words were extracted and analysed using linear mixed-effects models examining effects of training and clinical status. Children with Autism improved in signalling sentence prominence using duration but not f0 and intensity. Variability suggests that certain acoustic cues are more challenging. Comparing to TD children's focus marking patterns, autistic children's variability may also stem from their ongoing prosodic profile development.