A Novel Three Stage Process for Palmprint Verification

Palmprint identification/verification is becoming increasingly popularity in recent years. The fundamental step in palmprint based biometric authentication system is in locating the region of interest (ROI) in palmprint images. In this paper, we propose a novel three stage process of palmprint verification. Stage one involves a new method of locating and segmenting the ROI of palmprint images and it is based on locating the corner points between middle finger & ring finger. Second stage is based on texture feature extraction of palmprint using log Gabor transform and in third stage, we present a novel scheme for palmprint classification using Gaussian mixture model (GMM) and independent component analysis (ICA) and we call this as ICA mixture model (ICA MM). The proposed algorithm for classification involve the combination of texture information with GMMand ICA. The implementation of the algorithm involve combining texture and appearance based method (log Gabor & ICA). The proposed approach is validated for their efficacy on polyU palmprint database of 350 users. The experimental results illustrates the effectiveness of our methods. The improvement in performance can also be attributed to effective ROI extraction prior to using these algorithms.