Smart Belt: Human Motion Prediction and Fall Prevention from Wearable Sensor
Dynamic bipedal walking is challenging to model as well as replicate on robotic platforms. Most exoskeletons use body sensors (surface EMG’s, subdermal, etc.) to predict leg motion, but the observation of the environment can reduce the need for invasive sensors and provide a longer prediction horizon. Can the observation of the surrounding traversable area be used to assist in step prediction? Can surrounding observation reduce the need for invasive sensors in prediction? Can the learned gait-surrounding relationship be used to a) assist exoskeleton motion b) provide learning by demonstration to bipedal robot c) be used by elderly and visually impaired to avoid falls?
Specifically for fall prevention for the elderly or visually impaired, our goal is to develop a wearable sensor that can predict a person's path and their stability over the expected path, alerting them if there is a significant possibility they may fall or become unstable.
We show the ability to predict multiple paths a person may take, along with the correlated likelihood of taking a given path
Current Students
User Studies
For current user study please visit this page Fall Prevention Sensor - User Study (ongoing)
Related Publications
As a first step, we show here the ability to collect data from the person and surroundings useful for path, gait, and stability prediction. This is done by torso-mounted Intel Realsense cameras, and inertial measurement units on the thigh and calf of the user all integrated for pose estimation.