Real-Time 3-D Human Body Tracking using Variable Length Markov Models

Abstract

In this paper, we introduce a 3-D human-body tracker capable of handling fast and complex motions in real-time. The parameter space, augmented with first order derivatives, is automatically partitioned into Gaussian clusters each representing an elementary motion: hypothesis propagation inside each cluster is therefore accurate and efficient. The transitions between clusters use the predictions of a Variable Length Markov Model which can explain highlevel behaviours over a long history. Using Monte-Carlo methods, evaluation of model candidates is critical for both speed and robustness. We present a new evaluation scheme based on volumetric reconstruction and blobs-fitting, where appearance models and image evidences are represented by Gaussian mixtures. We demonstrate the application of our tracker to long video sequences exhibiting rapid and diverse movements.

Voxel based annealed particle filtering for markerless 3D articulated motion capture

“…In this field, some contributions employed linear techniques such as Kalman filtering [5] although being prone to loose track. However, due to the multimodal shape of the involved distributions, techniques based on the particle filtering algorithm (PF) proved to be efficient to tackle this problem as done in [6]. Finally, state-of-the-art annealed particle filtering (APF) introduced by [7] in the field of HMC noticeably improved the performance and robustness of the obtained results.…”

Real-Time Human Pose Inference using Kernel Principal Component Pre-image Approximations

“…Caillette et al [5] can capture at speed of 10Hz, but requires volumetric reconstruction from four well calibrated cameras. Navaratnam et al [11] uses monocular sequences to capture the upper body at a rate of 0.7Hz.…”

“…In this paper, we concentrate on the problem of inferring articulated human pose from pre-processed silhouettes. Due to ambiguities of the 2D silhouette to 3D pose space mapping, previous silhouette based techniques, such as [2,5,6,7,9,10,11], involve complex and expensive schemes to disambiguate between the multi-valued silhouette and pose pairs. A major contributor to the expensive cost of human pose inference is the high dimensionality of the output space (composing of 19 joints or 57 degrees of freedom in our setting).…”

Feature-Based Annealing Particle Filter for Robust Body Pose Estimation

“…Other authors use Visual Hull approaches (Laurentini, 1994) to work with voxel data. In that case, they can use three-dimensional flesh models, like ellipsoids (Mikic, 2003) or three-dimensional Gaussian mixtures (Caillette et al, 2005).…”