Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Wrist movement classification for adaptive mobile phone based rehabilitation of children with motor skill impairments

Kayleigh Schoorl, Tamara Pinos Cisneros, Albert Ali Salah, Ben Schouten

TL;DR

This work addresses the challenge of keeping children with motor impairments engaged in hand therapy by deploying a mobile exergame that classifies wrist dorsiflexion using smartphone sensors. It evaluates a spectrum of classifiers, with a CNN on extracted time-series features delivering high accuracy while remaining suitable for real-time mobile use. An adaptive difficulty system uses range of motion and speed indicators to tailor tasks to individual users, including a pilot with healthy subjects and a CP cohort. The findings indicate that dorsiflexion can be reliably detected in real time, enabling personalized, motivating rehabilitation, though further data and longitudinal studies across devices are needed to generalize the results.

Abstract

Rehabilitation exercises performed by children with cerebral palsy are tedious and repetitive. To make them more engaging, we propose to use an exergame approach, where an adaptive application can help the child remain stimulated and interested during exercises. In this paper, we describe how the mobile phone sensors can be used to classify wrist movements of the user during the rehabilitation exercises to detect if the user is performing the correct exercise and illustrate the use of our approach in an actual mobile phone application. We also show how an adaptive difficulty system was added to the application to allow the system to adjust to the user. We present experimental results from a pilot with healthy subjects that were constrained to simulate restricted wrist movements, as well as from tests with a target group of children with cerebral palsy. Our results show that wrist movement classification is successfully achieved and results in improved interactions.

Wrist movement classification for adaptive mobile phone based rehabilitation of children with motor skill impairments

TL;DR

This work addresses the challenge of keeping children with motor impairments engaged in hand therapy by deploying a mobile exergame that classifies wrist dorsiflexion using smartphone sensors. It evaluates a spectrum of classifiers, with a CNN on extracted time-series features delivering high accuracy while remaining suitable for real-time mobile use. An adaptive difficulty system uses range of motion and speed indicators to tailor tasks to individual users, including a pilot with healthy subjects and a CP cohort. The findings indicate that dorsiflexion can be reliably detected in real time, enabling personalized, motivating rehabilitation, though further data and longitudinal studies across devices are needed to generalize the results.

Abstract

Rehabilitation exercises performed by children with cerebral palsy are tedious and repetitive. To make them more engaging, we propose to use an exergame approach, where an adaptive application can help the child remain stimulated and interested during exercises. In this paper, we describe how the mobile phone sensors can be used to classify wrist movements of the user during the rehabilitation exercises to detect if the user is performing the correct exercise and illustrate the use of our approach in an actual mobile phone application. We also show how an adaptive difficulty system was added to the application to allow the system to adjust to the user. We present experimental results from a pilot with healthy subjects that were constrained to simulate restricted wrist movements, as well as from tests with a target group of children with cerebral palsy. Our results show that wrist movement classification is successfully achieved and results in improved interactions.
Paper Structure (20 sections, 11 figures, 2 tables)

This paper contains 20 sections, 11 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Games targeted at children with motor skill impairments
  4. Human activity recognition using smartphone sensors
  5. Data collection
  6. Classification of Dorsiflexion
  7. Feature extraction and selection
  8. Classification
  9. Multilayer perceptron (MLP)
  10. Convolutional neural network (CNN)
  11. Long Short-Term Memory
  12. Bidirectional Long Short-Term Memory
  13. Experimental Results
  14. Adapting the Game to the User
  15. Rule-based adaptive difficulty
  16. ...and 5 more sections

Figures (11)

  • Figure 1: The monster case used for the Magic Monster application and a screenshot from the mobile app.
  • Figure 2: Recorded sensor data for movement class 1: dorsiflexion with the phone upright and the screen turned away from the hand palm.
  • Figure 3: Recorded sensor data for movement class 21: rotating the phone to the left and right side alternately, with the phone held horizontally and the screen turned upward.
  • Figure 4: Dorsiflexion while wearing an arm brace with a plastic ruler inserted.
  • Figure 5: Dorsiflexion while wearing an arm brace with a metal bar ruler inserted.
  • ...and 6 more figures