An Approach for the Evaluation of Human Activities in Physical Therapy Scenarios

Human activity recognition has been widely studied since the last decade in ambient intelligence scenarios. Remarkable progresses have been made in this domain, especially in research lines such as ambient assisted living, gesture recognition, behaviour detection and classification, etc. Most of the works in the literature focus on activity classification or recognition, prediction of future events, or anomaly detection and prevention. However, it is hard to find approaches that do not only recognize an activity, but also provide an evaluation of its performance according to an optimality criterion. This problem is of special interest in applications such as sports performance evaluation, physical therapy, etc. In this work, we address the problem of the evaluation of such human activities in monitored environments using depth sensors. In particular, we propose a system able to provide an automatic evaluation of the correctness in the performance of activities involving motion, and more specifically, diagnosis exercises in physical therapy.