PERSONAS ARTEFACTOS

In this chapter, we proposed DyAd, a system that dynamically adjusts the difficulty level of physical exercises performed with the Proficio robotic system and based on

quantitative performance measurements. We showed that DyAd recommends a more difficult arm exercise if the current exercise is not challenging enough; it recommends

an easier exercise if it discovers that the user has trouble following the current exercise.

The approach to use DTW to align time series of gestures has been stated before in

the literature. Novel contributions of our work are (1) applying DTW in conjunction

with the spectral arc length method to analyze trajectories and (2) designing a system

that dynamically adjusts the difficulty of exercise based on this analysis.

In the experiment reported in this chapter, users without disabilities tested the

efficacy of our proposed system DyAd. The experimental outcome encourages us to prepare for and conduct a study that includes therapists and their patients. Our

future work will also include the design of appropriate exercises for DyAd, as well as virtual interfaces that enable tele-rehabilitation. In addition, due to high cost of the

Chapter 4

ExerciseCheck: Remote Monitoring and

Evaluation Platform for Home Based

Physical Therapy

4.1

Introduction

Home-based exercising is a vital part of any physical therapy program. With correct

execution of the exercises, faster recovery from physical problems can be achieved.

With the conventional approaches, however, a home-based physical therapy program

may not be as effective due to the lack of supervision by the therapist at home. Exer-

ciseCheck is designed as a remote monitoring and evaluation platform for individuals

involved in home-based physical therapy. The goal of ExerciseCheck is to provide

patients and physical therapists with real-time visual feedback and quantitative anal-

ysis. In this chapter, we discuss different elements of ExerciseCheck including data

analysis, hardware components, user interface, architecture, and data management.

For our research project, we have used the Microsoft Kinect. The Kinect is capable

of Skeletal Tracking which allows it to recognize people and map up to 25 joints on

their bodies. ExerciseCheck uses the Skeletal Tracking feature of the Kinect to locate

4.1.1 Procedure: Physical Therapy with ExerciseCheck

The goal of ExerciseCheck is to make sure that the exercises are performed correctly

at home when there is no supervision from the physical therapist. To do so, Ex-

erciseCheck requires a reference, so that it can compare the exercises performed at

home with that. Since each patient has different physical capability and perform the

exercise according to their strength, it is important that each reference is personal-

ized. Furthermore, to ensure that the reference is performed correctly, the therapist

supervision is required for recording the reference exercise.

Here is how ExerciseCheck works: In the clinic, the physical therapist initially

prescribes an exercise and asks the patient to practice the exercise multiple times

while the therapist is observing the movements carefully. Once the therapist assesses

that the patient is capable of correctly performing the exercise, the patient is asked

to perform the exercise in front of the Kinect. At that time, ExerciseCheck records

a reference trajectory for this exercise. The reference trajectory consists of the tra-

jectories of specified body landmarks. Our platform is capable of recording multiple

reference trajectories so that different exercises can be represented.

When the patient repeats a given exercise at home, ExerciseCheck captures the

trajectory of the same landmarks for data analysis. The patient can see the trajec-

tories of the reference and current positions of landmarks during the exercise. This

feature helps users to observe their mistakes and motivates them to put more ef-

fort toward moving more similarly to the reference. After each trial, ExerciseCheck

presents the result of the evaluation to both the patient and the physical therapist.

The trajectories and results are also stored on the server, so that the therapist can

4.1.2 Design and Development of ExerciseCheck

ExerciseCheck has been designed in three phases, where in each phase different as-

pect of the system has been designed, developed, deployed and tested. In order to

ensure the applicability of the Kinect for our purpose, we first need to examine its

performance in capturing the movement trajectory of the user during the exercise.

We begin this chapter by exploring the accuracy of the Microsoft Kinect and its ro-

bustness to clutter and different orientations of the body with respect to the camera

for an upper body exercise. Then we will discuss the three phases of this project.

Phase 1: Building a prototype

• Designing the initial architecture and data storage on a server, • Implementing quantitative analysis including accuracy and speed, • Designing the initial graphical user interface.

• Experiment: two Physical therapists and two users with no physical disability evaluated our platform. This phase confirmed the capability of our system and

the great potential in building such platform.

Phase 2: Major development

• Major development in the system architecture, data storage, and graphical user interface.

• Data analysis part of the design is updated, where the speed analysis is revisited, and repetition counting and analysis of active range of motion are added.

Phase 3: Final updates

• System updates based on the patient’s feedback.

• System preparation for the final deployment of ExerciseCheck at patient’s home. Updates includes methods to stop the recording automatically and with mini-

mum added noise.

• Experiment: ExerciseCheck was tested at patient’s home.