Emotion Recognition for Children with Autism—A Comparative Analysis of Different Classifiers

Emotion recognition refers to the process of identifying and understanding human emotions. Various cues such as facial expressions, vocal intonation, body language, and physiological signals can be used to identify human emotions. Several applications of emotion recognition scatter in several fields from human-computer interaction to the education sector, healthcare, and market research toward designing assistive technology. Children with autism spectrum disorder (ASD) suffer from neurodevelopmental disorder and may experience a range of challenges in social interaction, communication, repetitive behaviors, emotion regulation, etc. The objective of the proposed work is to accurately identify the different emotions of children with ASD and to make it easier for ASD children and those around them to interact with one another. The proposed research work has employed different classification algorithms to recognize six emotions namely ‘Joy’, ‘Anger’, ‘Sadness’, ‘Fear’, ‘Surprise’, and ‘Neutral’ of autistic children. The classification algorithms are applied over facial images on two existing datasets and the process consists of face detection, facial landmark localization, landmark representation etc.