Performance evaluation of transform based feature extraction methods for identity authentication system using fingerprint matching

Fingerprint verification is an important biometric technique for personal identification. This paper presents a performance evaluation of different fingerprint feature extraction methods. Fingerprint matching scheme based on transform features, like DCT (Discrete Cosine Transform), FFT (Fast Fourier Transform) and DWT (Discrete wavelet transform), have been presented and compared. In the fingerprint recognition application utilizing more information other than minutiae is much helpful. The technique described here obviates the need for extracting minutiae points to match fingerprint images. The transform coefficients are used to obtain the feature vector in terms of standard deviation and energy. Matching is performed using fast Euclidean distance between two feature vectors. The algorithms have been tested on database available from University of Bologna. Our comparison shows that DCT and FFT yields better GAR (Genuine acceptance rate) at low FAR (False acceptance rate) with reduced computational complexity over existing DWT and Gabor based algorithms. Because of reduced computational complexity these algorithms can be easily implemented as an embedded automatic fingerprint identification system (AFIS).