Design and Implementation of Fingerprint Recognition Using Minutiae Disk Code (MDC)

• Main Authors: Chu,Tsai-Te, 朱財德
• Other Authors: Chiu,Ching-Te
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/80288648309945189899

碩士 === 國立清華大學 === 資訊工程學系 === 103 === Fingerprint is one of the reliable and unique biometric features for the application of identity secure. Nowadays, mobile devices are also equipped with fingerprint sensor. The accuracy and the response time are important for the users. There are many researches aimed at raise the accuracy and reduce the computation time of fingerprint recognition. Minutiae cylinder code (MCC) is a coding method to encode the local neighbor minutiae and apply the local and global matching for fingerprint comparison, which has great performance on fingerprint recognition. However, the computation time of the MCC is high. Therefore, we proposed a minutiae cubic structure code (MCSC) method to speed up the computation time by simplifying the complex computation of the MCC. Furthermore, we proposed a new disk structure to encode the local structure for each minutia. The proposed minutiae disk code (MDC) encodes the neighbor minutiae more efficiently and clearly illustrate the distribution of the neighbor minutiae. The MDC method reduces the computation time by having 280.08x speed up in MCC encoding part on Matlab platform. In addition, the proposed MDC approach has high distinguish ability of the 96.81% recognition rate on the FVC2000 and FVC2002 datasets. The hardware implementation of MDC applies parallel computing on the relaxation process of global matching to accelerate the fingerprint comparison time. The implementation can achieve the operating frequency of 111MHz, conducting one fingerprint comparison of 0.81ms, which can process 1234 fingerprint images per second with the image size of 255x255 and the maximum of 64 minutiae, and with the core area of 0.48mm2 under TSMC 90nm technology. The hardware implementation has 141.27x speed up than the MCC method.