基于特征层和二代曲波变换的多模生物特征融合识别方法

An Algorithm for Multimodal Biometric Recognition Based on Feature Level and the Second-Generation Curvelet Transform

针对单模生物特征识别方法在实际应用中存在识别正确率较低的问题,提出了一种基于特征层和二代曲波变换的单样本多模生物特征融合识别方法,其采用了2种生物特征:掌纹特征和人脸特征.将所有归一化后的学习样本图像和测试图像通过组合的快速离散曲波变换和小波变换进行分解,系数经组合和规范化处理后,在特征层实现融合,融合后的特征参数送入K-最近邻分类器进行分类,从而获得最终识别结果.在香港理工大学掌纹数据库和Ljubljana大学人脸数据库上的实验结果表明,所提方法在每个类别仅使用1个学习样本的情况下,其生物特征图像的最佳平均识别正确率达到92.40%,比单模人脸、单模掌纹识别方法的识别率分别提高了35.38%和8.92%.

A single sample biometric recognition approach is proposed based on the feature level and curvelet transform of the second-generation to improve the recognition rate of the single modal hiometric system in application. Two kinds of biometric features are used. These are the palm-print feature and the face feature. All image samples are normalized and decomposed using the combination of curvelet ＆ wavelet transform. Then the normalized curvelet ＆ wavelet-trans- formed face and palm-print features are combined at the feature fusion level. The K-NN classifier is used to determine the final biometric classification, and then the recognition results are reported. The experimental results show that the proposed approach has better performance than the single modal solution： the best average recognition rate is improved to 92.4%,and the recognition rate is improved by 35.38% and 8. 92% compared with single face feature and single palmprint feature respectively.