Automatic digital biometry analysis based on depth maps Miguel Reyes a,b, *, Albert Clape ´s a,b , Jose ´ Ramı ´rez c , Juan R. Revilla c , Sergio Escalera a,b a Dept. Matema `tica Aplicada i Ana `lisi, UB, Gran Via de les Corts Catalanes 585, 08007 Barcelona, Spain b Computer Vision Center, Campus UAB, Edifici O, 08193 Bellaterra, Barcelona, Spain c Instituto de Fisioterapia Global Mezie `res, Guillem Tell 27, 08006 Barcelona, Spain 1. Introduction World Health Organization has categorized disorders of the musculo-skeletal system as the main cause for absence from occupational work and one of the most important causes of disability in elders in the form of rheumatoid arthritis or osteoporosis. It is estimated that 80% of the world population will suffer from musculo-skeletal disorders or dysfunctions (MSDs) during his life. As a result, MSDs lead to considerable costs for public health systems [1]. The body posture evaluation of a subject manifests, in different degrees, his level of physic-anatomical health given the behavior of bone structures, and especially of the dorsal spine. For instance, common MSDs such as scoliosis, kyphosis, lordosis, arthropathy, or spinal pain show some of their symptoms through body posture. This requires the use of reliable, noninvasive, automatic, and easy to use tools for supporting diagnostic. However, given the articulated nature of the human body, the development of this kind of systems is still an open issue. Given the difficulty of finding a tool for measuring the posture of the human body at different configurations, digital biometry has become a very useful tool. Digital biometry is defined by the American Society of Anthropometric data as ‘‘the technology to obtain reliable information of the physical objects or the environment through the recording of images, its measurement or interpretation’’. The systems based on this technology are capable to estimate morphological or functional alterations, being a useful resource for health professionals. The diagnostic evaluation of the anomalies follows through a careful study of musculo-skeletal structure and receptorial aspects. These diagnostic tools are based on monitoring anthropo- metric relationships with validated accuracy [21,38,34,23,22,27– 29]. These kind of tools are minimally invasive and obtain good accuracy results in terms of precision but require an specific scene configuration, being necessary a camera calibration preprocessing due to use of two-dimensional cameras in different planes. Another common handicap of these systems, is their reduced portability to perform a custom analysis for the therapist. These systems have been built highly parameterized for a specific type of analysis. Most of these systems only treat specific areas of the body, primarily the spinal deformities [23,24,32,30,31]. The solution more frequently applied to measure body posture consists of the installation and alignment of multiple cameras, applying stereo vision methodologies [3,4]. This kind of system uses to be expensive and require specific and restricted illumination condi- tions. The main alternative is accelerometers [5], but these systems Computers in Industry 64 (2013) 1316–1325 A R T I C L E I N F O Article history: Received 30 August 2012 Received in revised form 10 March 2013 Accepted 4 April 2013 Available online 27 May 2013 Keywords: Multi-modal data fusion Depth maps Posture analysis Anthropometric data Musculo-skeletal disorders Gesture analysis A B S T R A C T World Health Organization estimates that 80% of the world population is affected by back-related disorders during his life. Current practices to analyze musculo-skeletal disorders (MSDs) are expensive, subjective, and invasive. In this work, we propose a tool for static body posture analysis and dynamic range of movement estimation of the skeleton joints based on 3D anthropometric information from multi-modal data. Given a set of keypoints, RGB and depth data are aligned, depth surface is reconstructed, keypoints are matched, and accurate measurements about posture and spinal curvature are computed. Given a set of joints, range of movement measurements is also obtained. Moreover, gesture recognition based on joint movements is performed to look for the correctness in the development of physical exercises. The system shows high precision and reliable measurements, being useful for posture reeducation purposes to prevent MSDs, as well as tracking the posture evolution of patients in rehabilitation treatments. ß 2013 Elsevier B.V. All rights reserved. * Corresponding author at: Dept. Matema ` tica Aplicada i Ana ` lisi, UB, Gran Via de les Corts Catalanes 585, 08007, Barcelona, Spain. Tel.: þ34 608324411. E-mail addresses: mreyes@cvc.uab.es (M. Reyes), aclapes@cvc.uab.es, aclapes@cvc.uab.cat (A. Clape ´ s), jramirez@csc.uic.es (J. Ramı ´rez), jrevilla@csc.uic.es (J.R. Revilla), sergio@maia.ub.es (S. Escalera). Contents lists available at SciVerse ScienceDirect Computers in Industry jo ur n al ho m epag e: ww w.els evier .c om /lo cat e/co mp in d 0166-3615/$ – see front matter ß 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.compind.2013.04.009