A Digital Health Framework for Continuous Activity Monitoring, Emotion Recognition and Behavior Prediction in Children with Autism Using Wearable Iot and Artificial Intelligence

Abstract

Treatment of autism spectrum disorder (ASD) is challenging particularly as it pertains to continual expression of behaviour or emotions. The traditional clinical evaluation is periodic and subjective and is not able to give timely inputs for the interventions. In this study, a novel concept on “digital health” has been proposed which aims to achieve objective and continuous effective, automated monitoring in the field of autism using wearables IoT (IoT Healthcare Devices) and advanced Artificial Intelligence techniques. The framework also handles heart rate, electrodermal, temperature and movement to perform highly-comprehensive behavioural analysis. Three Al models (Random Forest (RF), Support Vector Machine (SVM) and Long Short-Term Memory (LSTM)) were used to determine the models' predictive ability. Results prove that the LSTM model outperforms the RF (89%) and SVM (85%) models in recognition behavior with sequential data having highest accuracy of 93%. With high precision, recall, and F1-score, the framework shows its consistency, allowing real-time monitoring. These results demonstrate the potential of wearable technology and deep learning to help provide data for personalized interventions that will optimise care and developmental outcomes for children with autism, provided in advance and by caregivers and health care workers.