基于GMM-WSUM的多生物特征二级融合识别方法-计算机工程与应用.PDFVIP

Computer Engineering and Applications 计算机工程与应用 2014 ，50（2 ） 179 基于GMM-WSUM 的多生物特征二级融合识别方法 1，2 1，2 李 雄 ，张东波 LI Xiong1，2 , ZHANG Dongbo1，2 1.湘潭大学 信息工程学院，湖南 湘潭 411105 2.智能计算与信息处理教育部重点实验室，湖南 湘潭 411105 1.College of Information Engineering, Xiangtan University, Xiangtan, Hunan 411105, China 2.Key Laboratory of Intelligent Computing Information Processing of Ministry of Education, Xiangtan, Hunan 411105, China LI Xiong, ZHANG Dongbo. GMM-WSUM based multiple biometrics identification with two level fusion strategy. Computer Engineering and Applications, 2014, 50 （2 ）：179-182. Abstract ：By combining the merits of density estimation and score normalization, a multi-biometric two level fusion method based on GMM （Guassian Mixture Model ）and WSUM （Weighted Sums ）, is proposed in this paper. N-P criteria is adopted as the first-level fusion strategy after the probability distribution of scores has been built by GMM model. And a weighted sums normalization is introduced as the second-level fusion strategy. This two level fusion method provides a better solution for the poor performance of score normalization fusion method when single-mode identification algorithms have great difference. The ORL and AR face database, FVC2004 fingerprint database are collected to build a face-fingerprint multimode database to evaluate the proposed method. Experimental results show that the proposed method effectively improve the performance of recognition. Key words ：Guassian Mixture Model（GMM ）; multi-biometric fusion; face; fingerprint 摘 要：结合基于密度估计和归一化两种融合方法的优点，在匹配分数层级提出了一种基于高斯混合模型（Guassian Mixture Model ，GMM ）和加权和（Weighted Sums ，WSUM ）的多生物特征二级融合识别方法。利用GMM 对匹配分 数建模后，采用N-P 准则作为第一级融合策略；第二级融合采用基于加权和的归一化方法，较好地解决了分数归一 化融合方法在单模识别算法识别率相差较大时融合识别性能差的问题。在ORL 、AR 人脸数据库和FVC2004 组成 的人脸-指纹多模数据库上进行了实验，结果表明，该方法有效地提升了识别性能。 关键词：高斯混合模型；多生物特征融合；人脸；指纹 文献标志码：A 中图分类号：TP391.4