An intelligent monitoring system for indoor safety of individuals suffering from Autism Spectrum Disorder (ASD)

• Published in: Journal of ambient intelligence and humanized computing Vol. 14; no. 12; pp. 15793 - 15808
• Main Authors: Manocha, Ankush, Singh, Ramandeep
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2023; Springer Nature B.V
• Subjects: Artificial Intelligence; Autism; Communication; Computational Intelligence; Data processing; Decision making; Deep learning; Engineering; Exercise; Feature maps; Indoor environments; Irregularities; Measurement techniques; Monitoring; Monitoring systems; Neural networks; Original Research; Questionnaires; Robotics and Automation; Safety measures; Sensors; Telemedicine; Three dimensional models; User Interfaces and Human Computer Interaction; Deep learning; Computer vision; Autism Spectrum Disorder (ASD); Smart surveillance; Therapeutic assistance
• DOI: 10.1007/s12652-019-01277-3
• ISSN: 1868-5137

Autism Spectrum Disorder (ASD) is distinguished by a variety of behavioral and social deficits. One of the most prominent problems in an autistic child is their aggressive nature that can cause physical injuries. In order to address the following behavioral deficits, this paper is an attempt to present a novel monitoring framework to predict irregularities from the physical activities of an autistic child. The primary motive behind this study is to improve the safety measures for the child in an indoor environment by generating time sensitive-alerts for caretakers and doctors. In this study, a deep 3D CNN and LSTM based activity prediction methodology are proposed to recognize physical irregularities. The 3D CNN model extracts the spatio-temporal features from the video templates for stance prediction with subject position. The LSTM model calculates the temporal relationship in the feature maps to analyze the scale of irregularity. Moreover, in order to deal with such irregularities, a time-sensitive alert-based decision process is proposed in the present work to generate early warnings to the doctor and caretaker. The proficiency of the system is increased by storing the performed activity scores in the local database of the system which can be further utilized to provide medicinal or therapeutic assistance. In addition, the measured experimental results are validated with state-of-the-art methodologies. Hence, the calculated statistical measures justify the utility of the proposed study in the healthcare and assistive-care domain.