Detecting stereotypical motor movements in the classroom using accelerometry and pattern recognition algorithms

Individuals with Autism Spectrum Disorders (ASD) frequently engage in stereotyped and repetitive motor movements. Automatically detecting these movements using comfortable, miniature wireless sensors could advance autism research and enable new intervention tools for the classroom that help children and their caregivers monitor, understand, and cope with this potentially problematic class of behavior. We present activity recognition results for stereotypical hand flapping and body rocking using accelerometer data collected wirelessly from six children with ASD repeatedly observed by experts in real classroom settings. An overall recognition accuracy of 88.6% (TP: 0.85; FP: 0.08) was achieved using three sensors. We also present pilot work in which non-experts use software on mobile phones to annotate stereotypical motor movements for classifier training. Preliminary results indicate that non-expert annotations for training can be as effective as expert annotations. Challenges encountered when applying machine learning to this domain, as well as implications for the development of real-time classroom interventions and research tools are discussed.