Ross Yeager

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2013-91

May 17, 2013

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2013/EECS-2013-91.pdf

Modern devices are integrating natural user interfaces (NUI) into their core structure through touchscreens, 3D depth measurement, voice recognition, and gesture recognition. The NUIs allow users to interact with their electronic devices in a manner that mimics their interaction with the physical world. Robust applications are needed to fully utilize the new demands and opportunities presented by such technologies. Because of the nature of NUI devices, there are opportunities to develop applications that support a wide range of people's physical activities. One such application is in remote physiotherapy, where patients use an at-home application to perform and track progress in a specified physical therapy program. This inherently involves the creation and utilization of a pre-defined exercise database that patients/therapists can compile to create custom rehabilitation regimens. Accordingly, there are challenges presented in defining and creating a relevant collection of these exercises. The mKinect project studies such a physiotherapy application based on Microsoft Kinect technology. This paper specifically focuses on the exercise definition required to support such an application. There are two categories of exercise definition: symbolic specification and learned demonstration. This project examines the combination of these categories in a hybrid manner, using both symbolic specification and a flexible exercise recording system to create exercises that can be used in the physiotherapy application itself.

Our system records joint angle, locations and limb lengths for key exercise poses from trainer demonstrations. Using quaternion interpolation, these key poses are then expanded into motion trajectories that can be tracked in realtime when patients perform these exercises. Variables include joint spatial positioning, joint angles, limb lengths, and hierarchical quaternion rotation angles. By automating the exercise definition process, therapists can rapidly add new exercises to the application database, providing a more practical and useful tool for therapists and patients alike. These exercises are then presented in XML format to an authoring user interface that allows the user/therapist to combine and sort exercises in the desired manner, allowing for more flexibility in patient therapy.

Advisors: Anant Sahai

---

BibTeX citation:

@mastersthesis{Yeager:EECS-2013-91,
    Author= {Yeager, Ross},
    Title= {An Automated Physiotherapy Exercise Generator},
    School= {EECS Department, University of California, Berkeley},
    Year= {2013},
    Month= {May},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2013/EECS-2013-91.html},
    Number= {UCB/EECS-2013-91},
    Abstract= {Modern devices are integrating natural user interfaces (NUI) into their core structure through touchscreens, 3D depth measurement, voice recognition, and gesture recognition.  The NUIs allow users to interact with their electronic devices in a manner that mimics their interaction with the physical world.  Robust applications are needed to fully utilize the new demands and opportunities presented by such technologies.  Because of the nature of NUI devices, there are opportunities to develop applications that support a wide range of people's physical activities.  One such application is in remote physiotherapy, where patients use an at-home application to perform and track progress in a specified physical therapy program.  This inherently involves the creation and utilization of a pre-defined exercise database that patients/therapists can compile to create custom rehabilitation regimens.  Accordingly, there are challenges presented in defining and creating a relevant collection of these exercises.  
The mKinect project studies such a physiotherapy application based on Microsoft Kinect technology.  
This paper specifically focuses on the exercise definition required to support such an application.  
There are two categories of exercise definition: symbolic specification and learned demonstration.  
This project examines the combination of these categories in a hybrid manner, using both symbolic specification and a flexible exercise recording system to create exercises that can be used in the physiotherapy application itself.  

Our system records joint angle, locations and limb lengths for key exercise poses from trainer demonstrations.  Using quaternion interpolation, these key poses are then expanded into motion trajectories that can be tracked in realtime when patients perform these exercises.  Variables include joint spatial positioning, joint angles, limb lengths, and hierarchical quaternion rotation angles.  By automating the exercise definition process, therapists can rapidly add new exercises to the application database, providing a more practical and useful tool for therapists and patients alike.  These exercises are then presented in XML format to an authoring user interface that allows the user/therapist to combine and sort exercises in the desired manner, allowing for more flexibility in patient therapy.},
}

EndNote citation:

%0 Thesis
%A Yeager, Ross 
%T An Automated Physiotherapy Exercise Generator
%I EECS Department, University of California, Berkeley
%D 2013
%8 May 17
%@ UCB/EECS-2013-91
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2013/EECS-2013-91.html
%F Yeager:EECS-2013-91