The Research Of Multimodal Biometric Recognition Based On Single Image

Biometric recognition is a field of technology that uses automated methods for identifying or verifying a person based on physiological or behavioral traits, has promising application prospect. Most of the current applications of biometrics recognition are based on unimodal biometric which has some limitations on the recognition rate and anti-spoofing. To improve the recognition rate and application scope of unimodal biometric systems, this paper take the sociality and openness of face feature and the high recognition rate of palmprint feature into consideration, proposed an identity authentication framework for the fusion of human face and palmprint based on a single image, and the fusion strategy have been investigated deeply. In this paper, the main works are as follows:(1) On review of the research history of biometrics and recognition technology, introduce 10 kinds of most common biometrics. Illustrate the goal and content of this paper by discussing the limitation of unimodal biometric systems, and also summarize the mainstream algorithms for the fusion levels.(2) Create the first multimodal database (face+palmprint) in this field which contains both face and palmprint information in one image. Avoid the phenomenon of extracting face and palmprint feature from different and independent database due to the lack of multimodal biometric database.(3) 11 kinds of local pattern is discussed in detail and have been tested on three major face database. Contrast and analyze the test results.(4) We proposed a novel local descriptor named Line based Weber Local Descriptor (LWLD). We can get the line response map and orientation map by modified finite radon transform (MFRAT). Calculate the Differential Excitation on the line response map and use the orientation map to replace the gradient direction information which calculated by original WLD. It has been validated that the proposed descriptor is effective in palmprint recognition by the experiments.(5) We proposed an identity authentication framework for the fusion of human face and palmprint based on a single image, which contains the following modules:face and palmprint detection, normalization and fusion. Verify the effectiveness of fusion by the experiments.