Recently, palmprint identification has been developed for security purpose. In this paper, we propose a novel palmprint recognition scheme which has three features: 1) representation of palmprint images by Local Binary Pattern (LBP); 2) dimensionality reduction by tensor subspace learning; and 3) recognition by multiple kernel classification method based on tensor analysis. LBP can effectively capture substantial palm features while keeping robustness to illumination. Then we reduce the dimensionality of each palmprint samples based on tensor subspace learning which can preserve the spatial structure of LBP. Tensor multiple kernel SVM classifier is finally employed for palmprint recognition. Experimental results on PolyU palmprint database show the effectiveness of the proposed method.