Robust arbitrary view gait recognition based on parametric 3D human body reconstruction and virtual posture synthesis

Abstract

Copies of full items can be used for personal research or study, educational, or not-for-profit purposes without prior permission or charge. Provided that the authors, title and full bibliographic details are credited, a hyperlink and/or URL is given for the original metadata page and the content is not changed in any way. Abstract: This paper proposes an arbitrary view gait recognition method where the gait recognition is performed in 1 3-dimensional (3D) to be robust to variation in speed, inclined plane and clothing, and in the presence of a carried 2 item. 3D parametric gait models in a gait period are reconstructed by an optimized 3D human pose, shape and 3 simulated clothes estimation method using multiview gait silhouettes. The gait estimation involves morphing a new 4 subject with constant semantic constraints using silhouette cost function as observations. Using a clothes-independent 5 3D parametric gait model reconstruction method, gait models of different subjects with various postures in a cycle are 6 obtained and used as galleries to construct 3D gait dictionary. Using a carrying-items posture synthesized model, 7virtual gait models with different carrying-items postures are synthesized to further construct an over-complete 3D 8 gait dictionary. A self-occlusion optimized simultaneous sparse representation model is also introduced to achieve 9 high robustness in limited gait frames. Experimental analyses on CASIA B dataset and CMU MoBo dataset show a 10 significant performance gain in terms of accuracy and robustness. 11

Gait Recognition and Understanding Based on Hierarchical Temporal Memory Using 3D Gait Semantic Folding

“…In this paper, to effectively extract the semantic features of gait, the 3D body instances from the make-human system [28] are used. No more attention is paid to the construction and training of 3D parametric gait model as numerous works have been published in our papers [29,30]. For simplification, we define the deformation function as F de (•), the 3D body modelŶ with a pose parameter ψ and a shape parameter ψ represented by the equation below.…”

“…where X std is the standard 3D body model data, D p (ψ) denotes the pose deformation with 3D joint semantic data ψ ∈ R N p where N p defines the number of joint parameters, and D s (S) defines the shape deformation with shape semantic parameters S ∈ R N s , where N s denotes the number of shape parameters, i.e., gender, height, weight, head size, torso size, arm size, leg size, breast size, and hip size. The joints are derived from the skeleton of the CMU mocap [31] and are encoded in a bio-vision hierarchical data (BVH) format as in our previous work [29].…”

“…As investigated in Reference [29], a 2D to 3D body estimation method is already proposed based on an energy cost function related to a 2D-3D silhouette difference. The binary 2D gait silhouettes are used for 3D body estimation in most of the previous research.…”

DeepGait: A Learning Deep Convolutional Representation for View-Invariant Gait Recognition Using Joint Bayesian

“…In general, video sensor-based gait recognition methods are divided into two families: appearance-based [3][4][5][6][7] and model-based [8][9][10]. Appearance-based methods focus on the motion of human body and usually operate on silhouettes of gait.…”

Decision-level fusion for single-view gait recognition with various carrying and clothing conditions

“…Recently, Guasch et al [18] used radar techniques to create a model of the human body based on the Doppler frequency 2 (Doppler Signature). Luo et al [54] used 3-dimensional parametric gait model reconstructed from multi-view silhouettes. 3D gait models were also used by Fernandez et al in [62] combined with 2D silhouettes analysis techniques.…”