基于改进ENN算法的多生物特征 融合的身份验证

Multi-Modal Data Fusion for Person Authentication Based on Improved ENN

基于多生物特征的身份鉴别技术已受到越来越多的重视.单个生物特征有其固有的局限性,通过融合不同的生物特征可以提高身份鉴别系统的验证性能和鲁棒性.该文融合了声纹和指纹特征,提出了一种改进的ENN方法,并与K-NN、传统ENN方法进行了比较.改进的ENN将认证率提高了大约2％.同时,又在不同的数据集上比较了改进的ENN方法和基于Bayes理论的融合系统,分析并评价了两种方法的适用范围和优缺点.实验结果证明了此方法的有效性.

People have paid more attention to biometrics based personal authentication recently. A single modality has its limitation in performance, such as universality and accuracy. The use of multiple modalities can get higher accuracy and winder universality. An improved ENN (Nearest-Neighbor with class Exemplars) method is proposed to fuse fingerprint and voiceprint. Compared with the traditional ENN and KNN (K-Nearest-Neighbor), the proposed method obtains further improvement of verification rates. The proposed method is also compared with the Bayesian fusion method. Performance of these two types of systems is given. Experimental results verify the validity of the proposed algorithm.