Training-based on real-time motion evaluation for functional rehabilitation in virtual environment

Abstract

One of the most effective applications of virtual reality (VR) in physical rehabilitation is training, where patients are trained for sequence decision-making in special situations presented in virtual environment. In this application, the evaluation of the movement of the subject performing a physical task is crucial. A good evaluation of the motion is necessary to follow the progression of the patient during his training session. Therefore, it helps therapist to better supervise therapeutic planning. Actually, the performance of the patient's training is determined by subjective observation of the therapist. Our approach is to propose a system that allows the patient to perform his training and to evaluate the progress of training in an autonomous way. This system consists of a motion analysis technique for a rehabilitation application where the patient is represented by his own avatar in virtual environment. The task performance required from the patient is his capability to reproduce in real time a movement. The real-time motion evaluation technique is based on the time series data matching method called Longest Common Sub-Sequence (hereafter LCSS). It is used to calculate distance between the reference motion of virtual avatar and the captured motion data of the patients and thus is used to determine how well the patients are doing during the training. The complexity of the technique proposed is in the order of O(δ) in which δ is a constant matching window size. Our prototype application is based on Tai-chi movements which have shown many health benefits and are increasingly used for therapeutic purposes.