A modified algorithm for fingerprint features extraction

Features extraction is crucial for fingerprint recognition, which reflects the superiority and decides the results of algorithm largely. An improved algorithm of fingerprint features extraction is proposed using wavelet transform and coefficient of variation. The coefficients are collected from the different scales and orientations in wavelet transform, and restrained the scales discrepancy owing to the coefficient, which strengthen the discrimination of eigenvectors. The k-nearest neighbor (k-NN) recognition is adopted in the simulation of the proposed method. The recognition rates of modified algorithm are much higher, which shows that the proposed method is effective in terms of a small-scale fingerprint recognition database.