Deakin University
Browse

File(s) under permanent embargo

Robust real-time bio-kinematic movement tracking using multiple kinects for tele-rehabilitation

journal contribution
posted on 2016-03-01, 00:00 authored by Pubudu PathiranaPubudu Pathirana, S Li, Hieu TrinhHieu Trinh, A Seneviratne
This paper investigates the robust and accurate capture of human joint poses and bio-kinematic movements for exercise monitoring in real-time tele-rehabilitation applications. Recently developed model-based estimation ideas are used to improve the accuracy, robustness, and real-time characteristics considered vital for applications, where affordability and domestic use are the primary focus. We use the spatial diversity of the arbitrarily positioned Microsoft Kinect receivers to improve the reliability and promote the uptake of the concept. The skeleton-based information is fused to enhance accuracy and robustness, critical for biomedical applications. A specific version of a robust Kalman filter (KF) in a linear framework is employed to ensure superior estimator convergence and real-time use, compared to other commonly used filters. The algorithmic development was conducted in a generic form and computer simulations were conducted to verify our assertions. Hardware implementations were carried out to test the viability of the proposed state estimator in terms of the core requirements of reliability, accuracy, and real-time use. Performance of the overall system implemented in an information fusion context was evaluated against the commercially available and industry standard Vicon system for different exercise routines, producing comparable results with much less infrastructure and financial investment.

History

Journal

IEEE transactions on industrial electronics

Volume

63

Pagination

1822-1833

Location

Piscataway, N.J.

ISSN

0278-0046

Language

eng

Publication classification

C Journal article, C1 Refereed article in a scholarly journal

Copyright notice

2016, IEEE

Issue

3

Publisher

IEEE