Abstract
Body-worn inertial sensors have enabled motion capture outside of the laboratory setting. In this work, an inertial measurement unit was attached to the upper arm to track and discriminate between shoulder motion gestures in order to help prevent shoulder over-use injuries in athletics through real-time preventative feedback. We present a detection and classification approach that can be used to count the number of times certain motion gestures occur. The application presented involves tracking baseball throws and volleyball serves, which are common overhead movements that can lead to shoulder and elbow overuse injuries. Eleven subjects are recruited to collect training, testing, and randomized validation data, which include throws, serves, and seven other exercises that serve as a large null class of similar movements, which is analogous to a realistic usage scenario and requires a robust estimator.