JoWUA

Abstract

Detecting autism spectrum disorder (ASD) is a difficult task for medical experts and healthcare providers, as the medical diagnosis is based primarily on the presence of a malfunction in brain functions that are reflected in the behavior of children. Facial expressions can be an effective alternative to diagnosing ASD. This is because children with ASD usually have specific patterns and behaviors that set them apart from other children. A key intervention enhanced the quality of medical home care for children on the autism spectrum by utilizing assistive technology. In this study, an emotion recognition system was developed for children with autism spectrum disorder. Face recognition, facial feature extraction, and feature classification are the three main stages of the proposed emotion recognition system. The system detects six facial emotions: fear, anger, sadness, neutral, surprise, and joy. The proposed system uses the mixing of Long Short-Term Memory (LSTM) and 3D convolutional neural network (3D-CNN) technologies to extract spatiotemporal features from pre-processed video templates. Findings reveal that the combination of the two techniques achieved an accuracy of 99.82%, which is better than other techniques.