In-Home Assistive System for Traumatic Brain Injury Patients

We describe a system for assisting patients in a home setting who suffer from cognitive impairments due to traumatic brain injury. The system integrates fixed wireless home sensors and wearable wireless sensors. We focus on the task of classifying activities of daily living. We locate and track the subjects with the help of home sensors and capture the details of an executed activity with a 2-axis wearable wireless accelerometer sensor attached to the right wrist. We extract time domain and frequency domain features for each task and classify them with Gaussian mixture models followed by a majority voter. The majority voter provides low false positive rates while continuously tracking the tasks. The experimental results from 2 subjects in recognizing 4 distinct daily activity tasks are promising.