Modeling Human Locomotion with Topologically Constrained Latent Variable Models

Abstract

Learned, activity-specific motion models are useful for human pose and motion estimation. Nevertheless, while the use of activity-specific models simplifies monocular tracking, it leaves open the larger issues of how one learns models for multiple activities or stylistic variations, and how such models can be combined with natural transitions between activities. This paper extends the Gaussian process latent variable model (GP-LVM) to address some of these issues. We introduce a new approach to constraining the latent space that we refer to as the locally-linear Gaussian process latent variable model (LL-GPLVM). The LL-GPLVM allows for an explicit prior over the latent configurations that aims to preserve local topological structure in the training data. We reduce the computational complexity of the GPLVM by adapting sparse Gaussian process regression methods to the GP-LVM. By incorporating sparsification, dynamics and back-constraints within the LL-GPLVM we develop a general framework for learning smooth latent models of different activities within a shared latent space, allowing the learning of specific topologies and transitions between different activities.