Autism Speech Therapy: A Novel RNN based Integrative Model for Improved Communication

Abstract

Children with Autism Spectrum Disorder (ASD) often face significant challenges in speech and communication. This study aims to develop a speech training model specifically designed for children with autism, integrating advanced machine learning techniques with therapeutic practices using mobile apps and store apps. The proposed model is designed to improve speech clarity in terms of diction and voice quality, and linguistic skills through customized and adaptive practising sessions. With the training data derived from the interactive app, the speech training model uses RNN and sequence of voice spectrum.   Preliminary results indicate that the proposed model achieves 86.6% accuracy for the raw data and 95.1% accuracy for processed data.