Recognizing body poses using multilinear analysis and semi-supervised learning

This paper presents an approach to full-body human pose recognition using features extracted from stereo silhouettes via multilinear analysis in a semi-supervised learning framework. Inputs to the proposed approach are pairs of silhouette images obtained from wide baseline binocular cameras. Through multilinear analysis, low dimensional view-invariant pose coefficient vectors can be extracted from these stereo silhouette pairs. Taking these pose coefficient vectors as features, a recently proposed state-of-the-art semi-supervised learning method, Universum, is adopted for pose recognition. Experiment results obtained using real image data showed the efficacy of the proposed approach.