Automatic newborn cry analysis: a non-invasive tool to help autism early diagnosis.

Autism Spectrum Disorders (ASD) are complex developmental disabilities that cause problems with social interaction and communication. ASD are associated with motor development problems, more or less blurred, and with perceptual and sensory brain areas. Crying is the infant's earliest form of communication and recent studies connect this original "language" form with autism disturbs. Being completely non-invasive, cry analysis is an appealing approach for early ASD diagnosis to improve rehabilitation. To this aim, we have developed an automatic system to record newborn cry and movements, during the first six months of life with a specific recording protocol. In this work we present first results of acoustic cry analysis in newborns classified as high-risk subjects being siblings of children already diagnosed as autistics. The work aims at finding possible early ASD signs in high-risk subjects as compared to a group of control subjects based on the fundamental frequency and the vocal tract resonance frequencies. Also, voiced and unvoiced parts of signal and cry-episodes duration are analyzed.