完备正交邻域保持判别嵌入的人脸识别

为解决邻域保持判别嵌入算法所面临的小样本问题，并充分利用类内邻域散度矩阵零空间和非零空间中的判别信息进行人脸识别,提出一种完备正交邻域保持判别嵌入的人脸识别算法。首先间接地利用特征分解方法去除总体邻域散度矩阵的零空间;然后分别在类内邻域散度矩阵零空间和非零空间中提取最优判别矢量。此外,为进一步提高算法的识别性能,给出了基于瘦QR分解的正交投影矩阵的求解方法。在ORL和Yale人脸库上验证了以上算法的有效性。

Face recognition based on complete orthogonal neighbourhood preserving discriminant embedding

In order to address Small Sample Size (SSS) problem encountered by Neighbourhood Preserving Discriminant Embedding (NPDE) and make full use of the discriminant information in the null space and non-null space of within-neighbourhood scatter matrix for face recognition, this paper proposed a Complete Orthogonal Neighbourhood Preserving Discriminant Embedding (CONPDE) algorithm for face recognition. The algorithm firstly removed the null space of the total neighbourhood scatter matrix using eigen decomposition method indirectly. Then, the optimal discriminant vectors were extracted in the null space and non-null space of within-neighbourhood scatter matrix, respectively. Besides, to further improve the recognition performance, the orthogonal projection matrix obtained based on economic QR decomposition was given. The experiments on ORL and Yale face database show the efficiency of the proposed method.