Person Validation System Using Hand Based Biometric Modalities

Abstract

Biometric system is useful for a pattern recognition system which makes a personal identification. Unimodal biometrics uses single evidence for identity e.g., gait, signature etc. There are some disadvantages faced by unimodal biometrics (such as accuracy, performance improvement etc.). So multimodal biometrics is proposed to overcome the disadvantages of unimodal biometrics. Multimodal biometrics uses multiple evidences e.g., palm print, knuckle print etc. This project proposes a multimodal biometric identification system based on inner knuckle print and palm print features of the human hand. The hand image is captured using digital camera. It is then pre-processed to get Region of Interest (ROI) of palm and knuckle. Palm print features are extracted from palm region using Histogram of Oriented Gradients (HOG) algorithm. The HOG algorithm is used because it exhibits the local object appearance and shape within an image. Also it can be used in the case of mehandi designs in the hand. Similarly the features from knuckle will be extracted using RIDGELET transform. It gives the efficient features in the ridges. Different features are extracted from palm print and inner knuckle print and these features are combined by using efficient fusion scheme i.e.) Discriminant Correlation Analysis (DCA) algorithm. And the final decision is taken by efficient classification module.