Fingerprint recognition using gabor filter with neural network

Abstract EN

The automated classification and matching of fingerprint images has been a challenging problem in pattern recognition over the past decades.

This paper proposes a method to detect the rotation region based on Estimate Global Region (EGR) that has the maximum rotation region.

The Gabor filter based feature is applied for extracting fingerprint features from gray level images without preprocessing.

The fingerprint recognition is developed by neural networks with adaptive learning rate.

The paper contains a comparison between using EGR algorithm and without using EGR.

The Gabor filter without EGR gives the best result for the fingerprint recognition with outrotation while the rotation of the fingerprint with angles (5o, 10o and 20o) gives worse results in fingerprint recognition.

The proposalmethod gives best result in rotation the fingerprint image with and without the rotation of the same angles.

The result of the correlation for the proposalmethod is 99 %.