基于广义典型相关分析融合和鲁棒概率协同表示的人脸指纹多模态识别

Multimodal Recognition of Faces and Fingerprints Based on Generalized Canonical Correlation Analysis and Robust Probabilistic Collaborative Representation

针对单模态生物特征识别容易受自身条件和环境变化的影响,鉴于人脸识别和指纹识别已经在生物识别系统中得到了广泛应用,提出了二者特征信息融合的多模态生物特征识别方法。该方法首先对人脸、指纹图像进行预处理,并对这两种模态均提取LBP和Gabor特征,然后将广义典型相关分析方法分别引入到人脸多特征融合和指纹多特征融合中,应用分块对角矩阵组合上述融合的人脸特征和指纹特征,最后用鲁棒概率协同表示分类器进行分类。在两个多模态数据库上的实验结果表明:与人脸或指纹单模态生物特征识别相比,基于人脸指纹的多模态生物特征识别具有更高的识别率和更好的稳定性;所提出的基于广义典型相关分析的特征融合方法优于传统的融合方法。

Unimodal biometrics is easily affected by conditions and environmental variations. Since face recognition and fingerprint recognition have been widely used in biometric systems, a multimodal biometrics method was presented based on the fusion of face and fingerprint features. The image processing methods were applied to the face and fingerprint images, and both LBP and Gabor features were extracted for these two biological features. Then, a generalized canonical correlation analysis was introduced into the multi-feature fusion of faces and multi-feature fusion of fingerprints respectively. A block diagonal matrix was taken to combine the above fused face features and fingerprint features.Finally, a classification was executed by solving the robust probabilistic collaborative representation.The experimental results on two multimodal databases demonstrate that the multimodal biometrics of faces and fingerprints has higher accuracy and better stability than the unimodal biometrics. The proposed fusion method based on the generalized canonical correlation analysis is superior to traditional methods.