Research On Identity Authentication Technology Based On Multi-modal Biometrics

The technology of identify authentication play a more and more important role in our life.There are so many ways of authentication which we often use. The one way that is used broadly is based on bio-characteristic authentication. What make it be applied broadly is that it has uniqueness, cannot copy and needn't remember. Recently, bio-characteristic authentication often have only one way to solve the problem. Although that is used convenience, it also have some disadvantages. For the recent problem, the research found that we can make two or more bio-characteristic fused to improve the robustness of recognition and safeness.The research of this paper is based on multi-modal fusion. It make the recognition of fingerprint and voiceprint fuse to authenticate identify. The content of this paper include:1. The research only is fingerprint recognition. The fingerprint which come from the machine to collect it cannot be used directly. Thus, before we abstract the characteristic of fingerprint, we need preprocess the photo of fingerprint. Then the detailing photo of fingerprint will be given. The next step is that the characteristic details will be extracted by the photo have been processed. The last step is recognized, which is based on the model of characteristic pattern. By the experiment,the result of equal error rate is 0.3678%.2. The research only is voiceprint recognition. It is similar to fingerprint recognition. Before the recognition, we need preprocess. Then we extract its feature MFCC. The next we use the GMM-UBM to recognize. By the experiment, the result of equal error rate is 0.4767%.3. The research is that fingerprint and voiceprint recognition fuse. On the level of fusion, this paper use different ways to fuse on the level of matching. The one is that use weighting fusion to make the recognition of fingerprint and voiceprint fuse together. The other way is applying the D-S evidence theory to fuse the feature. From the research, we found these two ways all improved the recognition rate, and it proved that more bio-characteristic fused can improve the recognition rate.