NSCTWavelet: An efficient method for multimodal biometric recognition based on pixel level fusion

Biometric recognition is emerging as a powerful means for automatically recognizing a person´s identity with a higher reliability. Especially, multimodal biometric fusion techniques have received a lot of attention and interest in recent years. To improve the recognition rate of the singlemodal biometric system and to solve the small samples recognition problem, a multimodal biometric recognition approach based on pixel level fusion is originally proposed in this paper. We use two kinds of biometric techniques: palmprint recognition and face recognition. Firstly, all image samples are normalized and decomposed using nonsubsampled contourlet transform. Then we combine the normalized nonsubsampled contourlet-transformed face and palmprint features at the pixel fusion level. At last, the KNN classifiers are used to determine the final biometric classification. and biometric recognition results will be obtained. The experiments are performed on the two well-known databases: CVL database and PolyU palmprint database. We only use one sample of every class as the training sample in the experiments. The experimental results indicate that the proposed approach has better performance than singlemodal solution. and the proposed approach can effectively extract the detailed feature information of the biometric images.