Autistic verbal behavior parameters

Abstract

The main goal of this paper is to present a method that allows the automatic detection of communication patterns of patients with a severe state of ASD (Autism Spectrum Disorder), through audio and video processing. Patients with ASD are have social and communicative disorders that make it difficult to analyze their expression. Their natural language presents different degrees of alteration, from a slight difficulty of expression to the total impossibility of speaking. The processing of the routines to be able to extract video and sound patterns is based on the work “ASD: ML perspective for Individual Performance Evaluation”. Taking the basis that stereotyped responses can be detected in a systematic way, similar to those that specialized therapists usually describe: gaze fixation, nodding, leg movements, clapping, among others; In this work, a model for the parameterization of said movements and sounds is proposed, through routines specially oriented to detect automatic patterns by audio and video processing. In this work there are the development and first results obtained from audio analysis of a use case, and main findings in relation on how the patient trend to interact with the environment. They allows to suppose that autistic verbal behavior might be interpreted by its physically manifested sounds.