A Pilot Study on Predicting Daytime Behavior & Sleep Quality in Children With ASD

Abstract

Sleep problems are a common concern for parents and caregivers in children with autism spectrum disorder (ASD). One of the main challenges in sleep studies with these individuals is the difficulty in monitoring sleep quality without sensors and wires attached to the subject’s body. Additionally, there is limited knowledge of how their sleep quality is related to their daytime behaviors. In this study, we evaluate an unobtrusive and inexpensive smart bed system for in-home, long-term sleep quality monitoring using ballistocardiogram (BCG) signals. By extracting different sleep quality indicators using BCG signals, we build bi-directional predictive models for daytime behaviors and nighttime sleep quality using two classifiers as support vector machine (SVM) and artificial neural network (ANN). For all daytime behaviors of interest, we achieve more than 78% average accuracy using previous nights sleep quality. Additionally, night time sleep qualities are predicted with more than 78% average accuracy using previous day and night features.