Application of SVM for evaluation of training performance in exergames for motion rehabilitation

Nowadays, the tools for remote monitoring and training analysis are a matter of deep interest in the field of telerehabilitation. In this study we present a method for the automated evaluation of performance in exergames for motor rehabilitation that can be performed by the patient, even autonomously in a domestic environment, with Microsoft Kinect and Leap Motion. The proposed method is based on a machine learning approach utilizing the Support Vector Machine (SVM). It uses a radial basis function kernel that deals with a two-class classification problem. The performance outcomes for one of the 10 exergames developed by our team are provided as a case study. After a crucial phase consisting of hyperparameter optimization, the SVM algorithm proved to be able to distinguish the "Good" class from the "Other" class with an accuracy of 0.80.