DEHFF – A hybrid approach based on distinctively encoded fingerprint features for live fingerprint detection

Abstract

Fingerprint-based biometric systems have elevated industrial standards worldwide through rapid personal verification. The array of identity frauds and cyber-attacks have also jeopardized the security countermeasures of biometric authentication systems. In this paper, an improved feature descriptor named Distinctively Encoded Histogram of Fingerprint Features (DEHFF) is proposed for live fingerprint detection. The DEHFF incorporates the two fundamental features i.e., ridge-valley contrast (ridge contours) and phase (ridge orientation) of fingerprints for liveness detection. The perceived response balances the abrupt intensity variations in the initial pixel information. Further, the Gabor filters are used to extract ridge contour information in the transform domain. Moreover, the objective function based on the Twin Confidence Value is proposed for linear integration of ridge contours extracted by Gabor and perceived spatial stimuli. The Local Phase Quantization utilizing the Point-wise Covariance of Gradients (PCG-LPQri) is proposed to determine the ridge orientation. Both feature sets are exclusively quantized into pre-defined intervals. The quantized features with relevant weights are then registered into a single feature set and represented through a 2-D histogram. Experimental evaluations indicate that the proposed DEHFF is able to reduce the average classification error to 4.01, 1.72, and 4.58% on LivDet 2011, 2013, and 2015, respectively.