Tensor space learning for analyzing activity patterns from video sequences

Abstract

Discovering knowledge from video data has recently attracted growing interest from vision researchers. In this paper, we describe a tensor space representation for analyzing human activity patterns in monocular videos. Given a set of moving silhouettes derived from raw video data, the proposed methodology first learns a tensor subspace model to embed the silhouettes into low-dimensional projection trajectories with preserved temporal order. Symmetric mean Hausdorff distance is then used to measure dissimilarity between the embedded motion trajectories in the tensor sub-space, as the basis for supervised or unsupervised learning. The experimental results on two recent video data sets have sho..