CMU-HCII-07-103
Human-Computer Interaction Institute
School of Computer Science, Carnegie Mellon University



CMU-HCII-07-103

Machine Perception for Occupational Therapy

Sonya J. Allin

August 2007

Ph.D. Thesis

CMU-HCII-07-103.pdf


Keywords: Machine Perception, Rehabilitation Engineering, Stroke Rehabilitation, Applied Computer Vision

In this thesis, we propose a novel measurement methodology for long term monitoring and assessment of stroke survivors°« functional motion. Our long-term goal is to measure and provide feedback about changes in functional movements that happen in everyday environments across days, months or years.

The research contained in the thesis consists of three parts. First, therapists were interviewed to determine common observations they make about stroke survivors functional activity for the purpose of assessment. Protocol analysis was used to gain insight onto parts of the body as well as motor symptoms that were particular points of focus.

Next, a prototype system was built to measure functional movement and its ability to capture statistics that discriminate between levels of functional impairment was tested. The technologies explored were primarily chosen because they are inexpensive and robust. They include simple computer vision and force sensing devices. Several different score prediction paradigms were tested, including some that are task specific and others that score movements, like lifts or grasps, across many tasks. We illustrate the accuracy of each paradigm, and discuss several of the underlying statistics found to correlate strongly and consistently with functional score. These statistics included the measured variance about the elbow and motion of the torso.

Finally, we present initial results from an effort to bring low cost and automated assessment technologies to the home. Cameras and force sensing devices were installed in the home of a single stroke survivor, and desktop activity was measured in this environment over the course of two weeks.

We believe that the technologies we highlight in this thesis have the capacity to inexpensively enhance the quality and character of therapy that stroke survivors receive after hospital discharge. This is primarily because they can focus therapy on real world tasks that take place in the home. We expect the tools we prototype here ultimately to enable: 1. Quantification of long term changes in functional mobility; 2. Evaluation of interventions that relate to functioning; 3. Motivating feedback about real world movement quality.

188 pages


Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu