Subject-specific Center of Mass Estimation for In-home Rehabilitation - Kinect-Wii board vs. Vicon-Force plate Alejandro Gonz´ alez, Mitsuhiro Hayashibe, and Philippe Fraisse Abstract— An increasingly aging society creates the need for a reliable evaluation of postural stability, specially for rehabilitation. Estimation of a subject’s center of mass (CoM) is important for the assessment of unsupported, stable standing. A portable, in-home estimation of CoM can be used as a rehabilitation tool and could be achieved using a Microsoft’s Kinect. To validate this approach we compare the performance of two statically equivalent serial chains. One of them was identified using a Kinect and a Wii board, while the other one was obtained from measurements performed with a motion capture system and a force plate. Their similar performance on a validation set indicates that it is feasible to perform subject- specific center of mass estimation in the home environment. I. INTRODUCTION Our society is aging rapidly. Currently, a little under 8% of the world’s population is 65 or older and this percentage is expected to reach 16% by 2050 [1]. The growth in the elderly population is more accentuated in developed countries where life expectancy continues to rise. As a result, the number of patients with motor function disorders can drastically increase while the ability to care for them will be limited by public expenditure and human resources. Thus, there is high demand for computer-aided tools which support in- home rehabilitation. Human static stability while standing is dependent on the relative position of the ground projection of the center of mass (CoM) to the support polygon described by the subject’s feet. The most common methods for CoM estimation require either knowledge of the subject’s pose, or measurement of his center of pressure (CoP) [2], [3]. In both cases high-end equipment is required. Such equipment is often expensive and can be accesible only inside the laboratory. The equipment set up is also complicated. Those constraints make in-home use unlikely. The statically equivalent serial chain (SESC) method simplifies calculation of the subject’s CoM. The SESC’s parameters can be identified from a set of known CoM ground projections, which are in turn estimated by CoP in the static case. In [4] a motion capture system was used to track subject motion while a force platform was used to provide CoP data. The need for a high-end motion capture system, limits the application of the method. A mobile measurement tool with a small set up time would make the SESC method available for wide scale application, especially for in-home rehabilitation. CoM based balance training for the elderly could be presented in the form of a game to enrich the user’s A. Gonz´ alez, M. Hayashibe and P. Fraisse are with INRIA DE- MAR Project and LIRMM, CNRS/University of Montpellier, France. {gonzalezde, hayashibe, fraisse}-at-lirmm.fr Fig. 1. Experimental setup for parallel measurement of Kinect-Wii board and Vicon-force plate. experience [5]. In [6] we demonstrated that it is possible to apply the method to data obtained with a Kinect sensor after identification performed using a Wii balance board. In this paper, we aim at comparing the result of CoM estimation using a SESC identified with a Kinect and a Wii board to a SESC obtained using a Vicon system and an AMTI-OR6 force plate. II. MATERIAL AND METHODS The Vicon system is capable of video-based motion tracking using the positions of several markers in 3D. Additionally, it deals with the synchronization of ana- logue data such as force plate information or EMG sig- nals. During identification, we measure the subject’s CoP using an AMTI-OR6 force plate. Parallel to the Vicon recording, we capture the subject’s movements using a Kinect to measure his pose and a Wii balance board for CoP information; (see Fig. 1). Processing of the Kinect data is done with the OpenNI-Primesense middleware (http://www.openni.org), while Wii board data is obtained using the Wiiuse project (http://github.com/rpavlik/wiiuse). In order to find the parameters of the SESC chain, healthy subjects were asked to hold a number of different static poses. Each pose was measured using 1) an 8 camera Vicon system with a marker placement which followed Plug-In- Gait recommendations and 2) a Kinect placed in front of the subject, 3 meters away. Ground reaction force data was measured for identification and validation purposes using 1) an AMTI-OR6 platform 2) a Wii board. To obtain in- formation from both devices at the same time, the Wii board was placed on top of the force plate during the test. Ground reaction forces were used to synchronize the Kinect and Vicon recordings. lirmm-00734972, version 1 - 5 Dec 2012 Author manuscript, published in "ICNR'12: International Conference on NeuroRehabilitation, Toledo : Spain (2012)"