Traditional person re-identification (Re-ID) methods usually use handc rafted features to describe pedestrian images.However,it is difficult to fully characterize the information of pedestrian images by single kind of features,r esulting in unsatisfactory recognition performance.A person re-identification m ethod that uses canonical correlation analysis (CCA) to fuse different kinds of features is proposed.First,the proposed method obtains the maximal correlative subspaces of two different types of visual features by using canonical correlat ion analysis.Then two fusion schemes,i.e.,the connecting fusion and the summi ng fusion,are applied to combine the two kinds of transformed features for pers on re-identification.Experimental results indicate that the proposed feature f u sion method can achieve better recognition performance while preserving lower fe ature dimensionality than the methods that uses single pedestrian feature and th e simple fusion strategy of directly connecting multiple types of pedestrian fea tures.