Metric for Exercise Recognition for Telemedicine Systems

The present paper has considered a task to develop quantitative metrics for automatic recognition of exercise types in the framework of the system for remote monitoring of the physical rehabilitation process. The paper aim is to develop metrics for estimation of physical exercise performance accuracy which allow obtaining quantitative characteristics of the treatment protocol fulfillment accuracy during rehabilitation exercise performance in order to increase objectivity and reliability of the remote control. The paper has suggested to use the algorithm of time scale warping with additional metrics localizing function properties to the minimum warping path. The scientific novelty: method for quantitative estimation of exercise performance correctness remotely based on the comparison of trajectories using the DWT metric has been developed. The present paper has suggested additional metrics that expand the DWT capabilities in the localization of trajectory changes.