Fingerprint Recognition Algorithm Design Based on Single Fingerprint Reference Data

Ng, Kah Chun (2018) Fingerprint Recognition Algorithm Design Based on Single Fingerprint Reference Data. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text NG KAH CHUN_full thesis.pdf Restricted to Registered users only Download (6MB)

Abstract

Fingerprint recognition is one of the common biometric technique use for personal identity. Fingerprint pattern is formed by ridges and furrows that construct the different type of minutiae. The fingerprint minutiae are unique for each person and it is used for verification and identification. The goal of this project is to design a fingerprint recognition algorithm by using the neural network, the single fingerprint reference will use as training data of the neural network. The concept of this algorithm is used as the fingerprint recognition application available now. For first registering of application such as the smartphone, security door and bank, user need put he or her fingerprint on the fingerprint sensor and scan one time. After registration success, the user can use the application for next times. The performance of proposed algorithm will compare with conventional algorithm. For the conventional algorithm, minutiae extractor and minutiae matcher technique were used. MATLAB software and both image processing and neural network toolbox are selected to achieve this algorithm. To obtain better minutiae extraction, pre-processing is executed for image enhancement and noise filtering. Some noise will cause the false minutiae marking, so false minutiae removal is required after minutiae marking. For minutia matching by the conventional algorithm, the similarity between two fingerprints is matched. Base on match score, it can differentiate the two fingerprints are same or not. For the proposed algorithm, single fingerprint reference will use as training data and the rest of fingerprints are used as testing data for classification. The false rejection rate (FRR) and false acceptance rate (FAR) were calculated and compare between proposed algorithm and conventional algorithm.