Recently, hyperspectral imaging has attracted more and more considerable research attention because of its discriminative information. This paper presents a multiple feature fusion strategy for hyperspectral palmprint recognition, combining texture LBP feature, direction LDP feature and global deep convolutional feature (DCNN). Compared to single patterns, the proposed method is focused on the combined use to improve the distinguishability. The LBP, LDP and DCNN are applied to the hyperspectral palmprint images (of an individual) in order to extract its features forming three feature matrixes. Each matrix contains redundant information and long dimensions. Thus, the 2D-PCA is applied to reduce the dimension of every feature matrix and generates a uniform feature vector. At last, residuals of each patterns based on the collaborative representation between the test data and training data are fused. The collaborative residual of the three patterns is exploited on final recognition. The proposed methodology was experimented on a large hyperspectral palmprint dataset consisting of 53 spectral bands obtaining an EER of 0.11% and an accuracy of 99.76% at recognition.