Abstract:

Aiming at the problem that the information of the face detection data is high eigenvector and face recognition is easily affected by expression changes, we propose a kernel principle component analysis (KPCA) face recognition method based on geodesic distance. We extract principal components by the nonlinear method. First, we adopt the KPCA method to map face data to the high dimensional space, and then principal components of the face are extracted in the high dimensional space, where the kernel function is a polynomial kernel. Finally, geodesic distance is introduced to replace the original Euclidean distance as the similarity measure, which can more accurately measure the actual distance between two points, and face recognition rate is less affected by expression changes as well. The proposed method cannot only achieve dimension reduction, but also achieve effective  data extraction. Experiments on the ORL face database show that the recognition rate of the proposed method is superior to that of the PCA and the KPCA.

Key words: face recognition, feature extraction, PCA, kernel function, geodesic distance