Multi-view gait recognition based on tensor analysis

Abstract

In this paper, we have proposed a tensor analysis method for multi-view gait recognition. Gait Energy Image (GEI) is used for gait feature and then 4-order tensor is composed. Using a generalized singular value decomposition method-HOSVD, this tensor is decomposed to person subspace, series subspace, view subspace and feature subspace. In person subspace we finish the multi-view gait recognition. Lastly, we perform recognition by using a simple nearest neighbor rule. Experimental results on the widely used CASIA-B gait database demonstrate the effectiveness of the proposed method.