Sparse Discrimination based Multiset Canonical Correlation Analysis for Multi-Feature Fusion and Recognition

Multiset canonical correlation analysis is a powerful technique for analyzing linear correlations among multiple representation data. However, it usually fails to discover the intrinsic sparse reconstructive relationship and discriminating structure of multiple data spaces in real-world applications. In this paper, by taking discriminative information of within-class and between-class sparse reconstruction into account, we propose a novel algorithm, called sparse discrimination based multiset canonical correlations (SDbMCCs), to explicitly consider both discriminative structure and sparse reconstructive relationship in multiple representation data. In addition to maximizing between-set cumulative correlations, SDbMCC minimizes within-class sparse reconstructive distances and maximizes between-class sparse reconstructive distances, simultaneously. The feasibility and effectiveness of the proposed method is verified on four popular databases (CMU PIE, ETH-80, AR and Extended Yale-B) with promising results.