Human Recognition Based on Gait Poses ⋆ Ra´ ul Mart´ ın-F´ elez, Ram´on A. Mollineda, and J. Salvador S´ anchez Institute of New Imaging Technologies and Dept. Llenguatges i Sistemes Inform`atics Universitat Jaume I. Av. Sos Baynat s/n, 12071, Castell´o de la Plana, Spain {martinr,mollined,sanchez}@uji.es Abstract. This paper introduces a new approach for gait analysis based on the Gait Energy Image (GEI). The main idea is to segment the gait cycle into some biomechanical poses, and to compute a particular GEI for each pose. Pose-based GEIs can better represent body parts and dynam- ics descriptors with respect to the usually blurred depiction provided by a general GEI. Gait classification is carried out by fusing separated pose- based decisions. Experiments on human identification prove the benefits of this new approach when compared to the original GEI method. Keywords: gait recognition, GEI, pose estimation, decision level fusion. 1 Introduction Gait can be defined as a manner or style of walking. Interestingly, there are studies asserting that every individual has a unique gait pattern [3], what has lead gait to be considered as a new biometric feature. When compared to other biometric features such as face, voice or fingerprint, gait has several attractive properties. It can be reliably perceived at a greater distance with simple instru- mentation, and it does not require the cooperation or awareness of the individual. There exist many applications that could benefit from gait analysis, including surveillance, diagnosis and treatment of gait-related disorders, motion capture in computer graphics and games, and so on. However, there are also several factors that hinder the use of gait as a biomet- ric feature. For instance, gait analysis is very sensitive to segmentation of the subject’s silhouette, but also footwear, clothing, carrying conditions and walk- ing speed may affect gait by reducing its discriminative power as a biometric. Even so, it is still useful to complement other biometric features under certain conditions (e.g. uncooperative subject, low quality images, etc.). In literature, two main approaches have been proposed to obtain gait patterns from video sequences [1]: model-based and model-free methods. Proposals in the first group aim at recovering a structural model of human motion [8,11] by ⋆ Partially supported by projects CSD2007-00018 and CICYT TIN2009-14205-C04-04 from the Spanish Ministry of Innovation and Science, P1-1B2009-04 from Fundaci´o Bancaixa and PREDOC/2008/04 grant from Universitat Jaume I. Portions of the research in this paper use the CASIA Gait Database collected by Institute of Au- tomation, Chinese Academy of Sciences. J. Vitri`a, J.M. Sanches, and M. Hern´andez (Eds.): IbPRIA 2011, LNCS 6669, pp. 347–354, 2011. c  Springer-Verlag Berlin Heidelberg 2011