Interactive Game-Based Motor Rehabilitation Using Hybrid Sensor Architecture

Interactive Game-Based Motor Rehabilitation Using Hybrid Sensor Architecture

Ahona Ghosh (Maulana Abul Kalam Azad University of Technology, India) and Sriparna Saha (Maulana Abul Kalam Azad University of Technology, West Bengal, India)
DOI: 10.4018/978-1-5225-9643-1.ch015

Abstract

Game consoles that use interactive interfaces have drawn users' attention as they reduce the total cost and are user-friendly too. This chapter introduces an interactive game to aid motor rehabilitation applicable to patients belonging to all age groups. In the system, the users receive some audio instructions regarding the action they have to perform next. According to that instruction, the user tries to complete the ‘Fruit-to-Basket' game as soon as possible by dropping all the fruits into the basket. Kinect sensor placed in front of the user detects their motions using skeletons containing three-dimensional coordinates of 20 body joints. The speeds of the movements are detected by the accelerometer. After extracting the required features from skeleton and speed, the use of principal component analysis is proved to be effective for feature space reduction. Then support vector machine is used efficiently to recognize the action. The experimental result indicates that the proposed algorithm is best suited in this domain and a very promising one.
Chapter Preview
Top

Introduction

Motor rehabilitation (Piscitelli, 2016) is the process to restore someone from physical weakness and disability through some training or therapy. The rapid development in sensor technology (Trankler & Kanoun, 2001) has made action identification presently an interesting research area. The application of video games as a restoration strategy (Bonnechère, Jansen, Omelina, & Van Sint, 2016) has immensely drawn attention of medical persons in the last decade and interactive game motivates one to be physically active without or with very less intervention of a trainer (Russell & Newton, 2008; Warburton et al., 2007).

Numerous works have been done in this domain. In (Schuldt, Laptev, & Caputo, 2004), authors introduced space-time feature to identify complex motion patterns with support vector machine (SVM). But it is limited to only some actions, while our proposed work deals with a larger set of actions. A serious game has been designed by Pedreza et al. (Pedraza-Hueso, Martín-Calzón, Díaz-Pernas, & Martínez-Zarzuela, 2015) for improving the physical and as well as cognitive functions of aged people. The first part of the game detects the player’s mobility, flexibility, strength and capacity where in the second part, the patient is asked to remember all the elements he/she had to face while playing the game. To overcome the problems of existing balance games like Nintendo Wii Fit, Lange et al. have proposed an iterative system to increase postural stability (Lange, Flynn, Proffitt, Chang, & “Skip” Rizzo,2010) and improve weight shift. The problem definition, discussion and refinement of ideas take place repetitively until the team decides the most appropriate idea. In (Nenonen et al., 2007), Nenonen et al. proposed a game, pulse master biathlon for skiing and shooting where the heartbeat of the subject controls the speed of skiing and with the increase of heart rate, the shooting screen gets faded, and so the player has to maintain a controlled heart rate to achieve good score in the game. Su et al. used dynamic time warping and fuzzy logic to design a home-based rehabilitation system (Su, 2013) where physicians get the summary report of exercise performed from cloud platform and prescribe medicine and exercise accordingly. Effects of virtual reality (VR) training have been investigated by Askin et al. for regaining of upper body functioning in stroke patients (Aşkın, Atar, Koçyiğit, & Tosun, 2018). Further studies having a higher number of patients need to be done to establish its effectiveness in neuro-rehabilitation.

This chapter presents a novel approach to motor rehabilitation by designing a strategy of score calculation while the game is being played by the subjects. The game is designed in such a fashion that the subjects are asked to perform certain tasks by moving their body parts based on specific audio commands. While doing so, the subjects’ body movements are measured using two distinct sensors, namely Kinect sensor (Lun & Zhao, 2015) (for action recognition) and Accelerometer (Chen, Liu, Jafari, & Kehtarnavaz, 2014; Tamura, 2014) (for speed calculation of body parts). We have incorporated these audio commands into a simple game of ‘Fruit-to-Basket’ in animated background. Here, the player has to collect different fruits according to some audio instructions and put them into the basket as fast as possible and also in an efficient manner (without colliding with obstacles). Game scores and performance details are stored in the computer and are analyzed after the completion of the game.

Complete Chapter List

Search this Book:
Reset