A Novel Fingerprint Matching Algorithm Based On Graphic Structures And Its Hardware Implementation.

With the development of our society and economics, automatic identification recogni-tion has been widely used in many fields, such as e-commerce and access control systems. The traditional methods(kn owl edge-based and token-based) have some security risks, such as copy and theft. Biometrics use individual's physical characteristics and behavioral char-acteristics to determine the identity of a person. Due to the uniqueness and the invariance of biological pattern, biometric identification methods are more effective than the traditional certification. Among many biometric technologies, fingerprint recognition is a better tech-nology, and automated fingerprint identification system has been widely in many fields of application.where fingerprint recognition is an important issue. It is believed that a large number of persons can be recognized, in part through fingerprint acquisition and analysis. In the early stage, fingerprint identification is mainly done by manual analysis of corresponding experts, such as forensics. However, it requires much time for the analysis of a single fingerprint, and if there are plentiful fingerprints or a large fingerprint database waiting for analysis, it will cost a plethora of time and human resources. Thence, an automatical fingerprint identification system(AFIS) can be used to improve the efficiency of fingerprint recognition.Though there are extensive studies about fingerprint recognition, the research in this field is not complete. The existed AFIS cannot reach some advanced requirements; they cannot achieve the very high recognition rate and the very low error rate, for example. Further improving the performance of the system has the important value in theoretical research and extensive application background. This paper mainly does some research about feature extraction and matching in fingerprint recognition, and accomplishes the following work:1. It organizes a detailed review about the main concepts and the usual technology about fingerprint recognition, and compares the different methods. Then the structure and the recent research are mentioned, and the proposed method is presented against some dis-advantages about feature extraction and matching in fingerprint recognition.2. A novel fingerprint matching algorithm is presented. It is based on a self-defined lo-cal structure, call waterstrider, which can extract the neighborhood information of minutiae. The information of neighborhood is the relationship between some minutiae and its con-nected minutiae. In the last, it uses a self-defined feature vector to present the feature from different samples. This feature vector is invariant to transformation and rotation. The com-parison between feature vectors can results in the output the decision. The experiments are conducted on FVC2002 database and the results illustrate the effectiveness of the proposed algorithm.3.A hardware implementation is presented, which is a DSP-based fingerprint recogni-tion system.