Computer vision based human computer interaction using color detection techni...
Kinect-VRehab
1. Kinect-VRehab: A Platform for Online and
Real-time Rehabilitation in Virtual-worlds with
Kinect
Candelario Guti´errez
Department of Computing Science
University of Alberta
Edmonton, AB. Canada
cgutierr@ualberta.ca
Abstract—We present Kinect-VRehab, an online virtual reha-
bilitation platform implemented using Microsoft Kinect. Kinect-
VRehab supports rehabilitation sessions using effective social
mechanics, movement-accuracy measurement algorithms, and
real-time on-screen feedback. Our tool is capable of comparing
patients’ motion with predefined exercise specifications provided
by physicians through a domain-specific language. In this paper,
we present a high-level overview of Kinect-VRehab’s architecture
and implementation.
I. INTRODUCTION
Many physical therapy exercises are designed to be executed
with the assistance rehabilitation medicine experts in a clinical
environment. This constrain limits the accessibility of the re-
habilitation spaces, given that patients often have to travel long
distances to the rehabilitation centres where specialized equip-
ment and personnel are located. Furthermore, more often than
not, patients and physicians suffer from scheduling problems
since different rehabilitation sessions may need simultaneous
access to limited rehabilitation resources. Virtual Reality (VR)
and motion capture technology can be used in order to make
rehabilitation spaces more accessible to patients, and more
scalable to physicians. They can potentially provide portable
tools that patients can use to complete rehabilitation sessions
from their place of residence, with the remote supervision of
rehabilitation professionals.
Some examples of virtual rehabilitation spaces include
Doctor Kinetic1
and KinectoTherapy2
. They both provide
interactive gameplay experiences, including, but not limited
to, target collection, and reaction-based games. However, these
environments are limited to isolated rehabilitation spaces,
where patients don’t have social interaction among each others.
Furthermore, they lack of real-time motion-accuracy feedback,
thus rehabilitation experts are recommended to be present
during the rehabilitation sessions to avoid patients’ injuries.
In this paper, we present Kinect-VRehab, an online virtual
rehabilitation platform implemented using Microsoft Kinect.
Kinect-VRehab supports rehabilitation sessions using effective
social mechanics, and precise feedback algorithms.
1http://doctorkinetic.nl/
2http://www.kinectotherapy.in/
Kinect-VRehab enables patients to perform exercise rou-
tines at any time of the day, and from the comfort of their
home or workspaces. Moreover, using Kinect-VRehab, patients
are able to connect to a VR rehabilitation world, join exercise
sessions with or without other patients, and experience indi-
vidual or social rehabilitation scenarios. Furthermore, Kinect-
VRehab provides a domain-specific language that allows
physicians to describe, from a high-level perspective, norma-
tive movements for exercise sessions. Such specifications can
be loaded to a patient’s profile, and then compared with the
patient’s motion execution. This comparison allows Kinect-
VRehab to provide real-time feedback using animated on-
screen indicators, thus correcting improper, and potentially
dangerous movements.
II. RELATED WORK
Let us now review some of the related work that uses
Microsoft Kinect and VR environments for rehabilitation pur-
poses. Fernando Cassola et al. [1] propose a prototype platform
called Online-Gym. It is focused in rehabilitation exercises for
people with reduced mobility. Using a Kinect and Online-Gym,
a patient can join a workout session in a VR world, where
a real instructor guides the session using an avatar. Chien-
Yen Chang et al. [2] compare the Kinect sensor with more
expensive alternatives, such as the OptiTrack3
framework, in
order to assess if Kinect is as precise and flexible as high-
end motion capture equipment in rehabilitation environments.
The results obtained by Chien-Yen Chang et al. showed that,
for the most common rehabilitation movements, the Kinect
sensors are as precise as OptiTrack.
III. METHODOLOGY
Kinect-VRehab is composed by three major components, a
local client application, a domain-specific language grammar
and parser, and an animation relay server based on OpenSim4
.
Figure 1 diagrammatically depicts Kinect-VRehab’s architec-
ture.
3https://www.naturalpoint.com/optitrack/
4http://opensimulator.org/
2. Fig. 1. Kinect-VRehab Software Architecture Overview
1. Local Application: The client application captures
the information from the Kinect sensors using Rinions5
a Real-time animator for OpenSim. These information is
sent to Rinions’s animation relay server which in turn uses
OpenSim in order to distribute the movement information
to all the OpenSim clients connected to the infrastructure.
Rinions enables Kinect-VRehab to automatically send and
receive animation files in a BVH format, so every participant
of a rehabilitation session has immediate feedback of the
movements that other patients are performing. Furthermore,
by comparing patient’s performance with the physician’s
recommended (normative) movements, the client application
analyses the patient’s motion accuracy. Figure 2 portraits how
the feedback is displayed to the patient using a three color
scale indicator.
2. Server: The application server contains Rinions’
animation relay server together with an instance of the
OpenSim server. These two components work together
to support Kinect-VRehab’s distributed animation process.
Furthermore, the server stores the physician’s specifications
for a patient exercise routine using a web database.
3. Domain-specific Language: The domain-specific lan-
guage has been designed using xText6
, a language workbench
that generates language editors with syntax coloring, content
assistance, and grammar validation. Kinect-VRehab’s language
captures, from a high-level perspective, the normative move-
5http://www.nsl.tuis.ac.jp/xoops/modules/xpwiki/?Rinions
6https://eclipse.org/Xtext/
ments that a patient should perform during the execution of
a rehabilitation session. It uses concepts of extremity position
such as rotation angles, and flexion and extension displacement
lengths.
Fig. 2. Kinect-VRehab Rehabilitation Session Example
IV. CONTRIBUTIONS
In this paper we presented Kinect-VRehab, an online virtual
rehabilitation platform implemented using Microsoft Kinect.
Kinect-VRehab supports rehabilitation sessions using effec-
tive social mechanics, movement-accuracy measurement algo-
rithms, and real-time on-screen feedback. Our proposal aims
at providing a flexible rehabilitation platform that patients can
use from the comfort of their homes. Furthermore, Kinect-
VRehab provides a virtual space where patients can interact
with each other, adding a social dimension to the rehabilitation
sessions. Kinect-VRehab also provides a domain-specific
language so physicians can specify the normative movements
of rehabilitation scenarios. These specifications are in turn
used to present on-screen real-time feedback to the patients,
thus minimizing the need of on-site supervision for low
risk exercises. We are currently working towards extending
the domain-specific language to support the definition of
more complex normative movements, and a web interface
for patients and physicians that visualizes available exercise
batteries, and progress and performance statistics.
REFERENCES
[1] F. Cassola, L. Morgado, F. de Carvalho, H. Paredes, B. Fonseca, and
P. Martins, “Online-gym: A 3d virtual gymnasium using kinect interac-
tion,” Procedia Technology, vol. 13, pp. 130–138, 2014.
[2] C.-Y. Chang, B. Lange, M. Zhang, S. Koenig, P. Requejo, N. Somboon,
A. A. Sawchuk, and A. A. Rizzo, “Towards pervasive physical reha-
bilitation using microsoft kinect,” in Pervasive Computing Technologies
for Healthcare (PervasiveHealth), 2012 6th International Conference on.
IEEE, 2012, pp. 159–162.