基于特征融合的多模态身份识别方法研究

Research on multimodal biometric authentication using feature level fusion

研究了多模态身份识别问题,结合人脸和掌纹两种不同生理特征,提出了基于特征融合的多模态身份识别方法。对人脸和掌纹图像分别进行Gabor小波、二维主元变换(2DPCA)提取图像特征,根据新的权重算法,结合两种模态的特征,利用最邻近分类器进行分类识别。在AMP、ORL人脸库和Poly-U掌纹图像库中的实验结果表明,两种模态的融合能更多地给出决策分析所需的特征信息相比传统的单一模态的人脸或掌纹识别具有较高的识别率,更具安全性和准确性。

Multimodal biometric authentication method is proposed, which combines the features of human faces and palmprints. Biometrics features are extracted using Gabor wavelet and two dimensional principal component analysis （2DPCA） techniques, and identification is carried out by the nearest neighbor classifier according to the combined biometric features of two modals and a new weighting strategy. The AMP, ORL and Poly-U databases are used as the test data in the experiments. Experimental results show that combination of two different modals can provide more authentification information, which generates higher security and more accuracy than the single model authentication.