Multi-Modal Biometrics Pixel Level Fusion and KPCA-RBF Feature Classification for Single Sample Recognition Problem

The single sample recognition problem is a difficult research topic in the field of biometrics, since very limited training samples and image discriminant information can be acquired. We propose a new multi-modal biometrics fusion approach to try to solve this problem, which uses face and palmprint biometrics. We combine the normalized Gaborface and Gaborpalm images in the pixel level, and present a Kernel PCA plus RBF classifier (KPRC) to classify the fused images. Testing on a large face database (Feret) and a large palmprint database, the experimental results demonstrate that the proposed pixel level fusion approach can significantly improve the recognition effects of single-modal biometrics. In addition, our approach is superior to a conventional decision level fusion method.