【作者】 王轶； 金立左； 潘泓； 袁晓辉；

【Author】 WANG Yi,JIN Li-zuo,PAN Hong,YUAN Xiao-hui (School of Automation,Southeast University,Nanjing Jiangsu 210096,China)

【机构】 东南大学自动化学院；

【摘要】 为了解决单一模式生物特征识别因其固有的局限性而难以满足实际应用需求的问题,采用基于人脸和指纹的多特征融合的识别方法,在匹配后的得分层进行信息融合,利用逻辑斯蒂(logistic)判别方法进行分类。并在此基础上,提出了一种排序层和得分层相结合的融合方法,采用树状识别流程,先进行指纹识别,选取匹配得分排在前N位的样本进行人脸识别,并引入加罚措施,再对这两个分类器排序加权后的得分值进行融合、分类。针对自行构建的生物特征库,分别进行了单一人脸识别和指纹识别、匹配得分层融合的多特征识别、结合排序层和得分层融合的多特征识别,实验结果表明改进后的方法不仅能取得比单一特征识别更高的识别率,而且较之只在匹配得分层融合方法的识别率亦有明显提高。更多还原

【Abstract】 In order to resolve the limitations of single biometrics identification in practice, multimode biometrics identification technology based on faces and fingerprints is adopted.Feature information is fused in matching score level and logistic discrimination is used in classification.Moreover,a fusion algorithm associating ordering level and score level is raised,in which tree-like process and weighting are utilized.Toward the biometric library built by ourselves,face identification,fingerprint identification,multimode identification fused in score level and multimode identification fused by the new method are realized.Experiment result indicates that the new method can receive the best identification rate.更多还原