Coupled principal component analysis based face recognition in heterogeneous sensor networks

• A framework for combining local and global features in HSN is proposed.
• We propose CPCA for face recognition in HSN.
• The CPCA could reduce the gap between heterogeneous sensor nodes.

In this paper, we construct heterogeneous sensor networks (HSN) for face recognition and propose a novel approach named coupled principal component analysis (CPCA) that uses a feature-based representation for heterogeneous face images. We first employ local binary patterns (LBP) to capture the local structure of face images, and then propose CPCA to obtain the global face information. The proposed CPCA could incorporate the information between heterogeneous feature spaces, and therefore it reduces the gap between face images captured from heterogeneous sensors in HSN. Finally, the spare representation is utilized for matching heterogeneous face images. The experimental results demonstrate that the proposed approach achieves better performance than the state-of-the-art approaches.