Physiotherapy guidance by motion analysis based on Hidden Markov Model

A system that would allow people who can´t spare time for physical therapy or who have to stay at home due to disabilities, to perform exercises in the comfort of their homes and independent of time restrictions shall contribute positively to their health. While such a system supports the patients with feedback, it can also enable the monitoring of patients by health personnel. In this study we acquired the body models of people performing physiotherapy motion by using Kinect depth camera, which has recently become a popular sensor for image processing application. We implemented an application that compares the performed motion with the pre-stored correct motion model and gives feed-back to the patient. We modeled the motions using various Hidden Markov Models (HMM) having three different topologies possessing ergodic, temporal and both ergodic and temporal features. We observed that trained HMM´s have modeled the motions successfully and have provided useful feedback.