Interactive system with speech therapy for children with level 01 autism spectrum disorder

Abstract

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition characterized by social communication challenges and repetitive behaviors. Early intervention, typically before age two or three, is crucial despite the lack of a cure. Specialized therapies enhance communication, social skills, and quality of life. Severity levels range from Level 1 (least severe) to Level 3 (most severe). Speech therapy is essential for improving communication skills, using techniques to enhance speech, language, voice, and fluency. Technological innovations like facial expression and speech recognition improve diagnosis and treatment outcomes, aiding societal inclusion. The research aims to design an interactive system tailored for speech therapy for Level 01 ASD children around three. Objectives include integrating facial expression and speech recognition to enhance therapy effectiveness and developing an interactive platform for seamless communication during sessions. This approach aims to improve communication, social interaction, and therapeutic outcomes. The system includes sound and interactive touch systems for image selection and a PC-based intelligent system with a mic and webcam for interaction. Facial expression identification, interactive system engagement, and speech recognition are manually evaluated. The conclusion is based on facial expression identification, interactive system engagement, and speech recognition results. Results indicate good progress at 80%-100%, moderate progress at 50%-80%, and poor progress below 50% for Level 01 ASD children. The pre-trained facial expression recognition module effectively identifies emotions such as sadness, surprise, anger, and neutral states, integrating them into memory classification. During therapy, the system assesses similarities between typical human emotions and those of ASD children, enhancing interaction dynamics. The module includes a Haar cascade for face detection and a CNN model for facial expression identification. In the initial stage, the system uses real-time speech clips to construct the CNN model for speech recognition, aiding interaction during cartoon image selection. The overall accuracy of the developed system is 92.5% for together with speech recognition, facial expression identification and interactive platform engagement. This integrated approach aims to manage ASD symptoms effectively, improve communication skills, and enhance the quality of life for affected individuals. It underscores the importance of early intervention and personalized therapeutic strategies in supporting individuals with ASD.