Analysis of Joints for Tracking Fitness and Monitoring Progress in Physiotherapy

This work proposes a framework to track the progress in angles and range of motion of joints in physiotherapy. Using a sensor, the 3D skeletal information of a subject undergoing the therapy is extracted. Using time-frequency features of the skeletal profile based on the approximation coefficients of the Discrete Wavelet Transform (DWT), the exercise the subject is engaged in is identified by a recurrent neural network in conjunction with long-and-short term memory. Subsequently, each instance of the exercise is segmented. The best of these is used as a reference and various instances of the exercise are compared against the reference for repeatability, fidelity, etc., to study muscle and/ or joint fatigue and progress. Finally, Joint performance analysis is carried out using metrics evaluated at the end of each engaging session. Experimental results demonstrate that with such progressive analysis, it is possible to quantify the performance through the course of the regimen.

This work has been accepted for oral presentation at IEEE International Conference on Signal and Image Processing Applications (ICSIPA 2019).

Collaborators - Piyush M. Surana, Rahul Ragesh, Dr. Gowri Srinivasa

Please find all the relevant resources below:

1. Paper.