WearSense: Detecting Autism Stereotypic Behaviors through Smartwatches

Area of Science:

• Biomedical Engineering
• Neurodevelopmental Disorders
• Wearable Technology

Background:

• Autism Spectrum Disorder (ASD) affects 1 in 68 children in the U.S., presenting unique behavioral challenges.
• ASD is linked to intellectual disability, social skill deficits, and physical health issues.
• Effective monitoring tools are crucial for supporting individuals with autism.

Purpose of the Study:

• To introduce WearSense, an Internet-of-Things (IoT) framework utilizing smartwatch sensors.
• To assess the feasibility of detecting specific stereotypic behaviors in children with autism using smartwatch data.
• To evaluate the accuracy of behavior detection for potential clinical and caregiving applications.

Main Methods:

• Recruited 14 subjects with autism to wear smartwatches.
• Collected accelerometer data from the wrist-worn smartwatches.
• Extracted 102 features from the three-axis accelerometer data for analysis.

Main Results:

• An ensemble of 40 decision trees achieved the highest classification accuracy of 94.6%.
• The study demonstrated the effectiveness of smartwatch sensors in identifying hand flapping, painting, and sibbing behaviors.
• High accuracy validates the quality of collected data and feature extraction techniques.

Conclusions:

• Smartwatch-based sensing offers a promising, non-invasive method for monitoring autistic behaviors.
• The WearSense framework facilitates cloud-based data analysis for comprehensive insights.
• This technology can empower parents, caregivers, and clinicians in making informed decisions for autism support.