Analyzing the potential of physiological signals for emotion detection in children with autism

Autism spectrum disorder (ASD) is a neurodevelopmental disorder that affects communication, social interaction, and behavior of an individual. Studies are being conducted to understand emotions in individuals with ASD. This research is centered on the emergence of physiological signals that are a valuable tool for emotion detection. By monitoring parameters such as heart rate, skin conductance, and oxygen saturation using non-intrusive sensors to detect and understand emotional states in real-time. Considering Ekman’s model, we took happy, sad, and fear emotions. This technology not only aids in the detection of emotional distress but also facilitates personalized strategies for supporting emotional well-being in individuals with autism. Dataset collected from different schools, rehabilitation center, and university from students, autistic individuals, and bachelor programs students. Different machine learning algorithms were trained and tested on the collected dataset. KNN showed the highest accuracy when compared to SVM and Naive Bayes. An interface was developed where the predicted emotions were shown along with the recommendations to deal with them. The study demonstrates that physiological signals such as pulse rate, GSR, and SpO2 are reliable indicators for emotion detection. By leveraging machine learning models, emotional states can be effectively identified.