Biometric Security System using the Inner Knuckles

Abstract

Biometric security systems are now highly explored by many researchers for their high accuracy levels. These systems were recorded to be more reliable than traditional security systems that use typed passwords, as they are robust against hacking. Some biometric prints have been explored to be used in security systems. These prints include face, iris, voice, palm, finger print, knuckles print, handwriting, speech and keystroke. The knuckle prints (skin patterns) which are formed at the joints either in the finger back surface (Outer Knuckles) or in the finger inner surface (Inner Knuckles) can be captured by contact or contactless devices and exhibit promising accuracy results. In this thesis, first an extensive review has been carried out and a comparison is done and presented between the performance of some of holistic-based and local feature-based recognition methods on a single print and on multiple prints fused together. The performed review also covered the available datasets that can be used for performance evaluation. Second, a personal recognition method using the Center Inner Knuckle Prints has been proposed. The proposed method uses the Neighboring Direction Indicator features along with a perfect alignment and enhancement preprocessing steps to boost the performance compared to state-of-the-art methods. The performance of the proposed method has been investigated using different databases composed of low resolution hand images captured by a contactless capture in a free environment: Sfax-Miracl Hand Database, Ground Truth of Sfax-Miracl Hand Database, Ground Truth of PolyU Contact-free 2D/3D Hand Images Database and Ground Truth of IIT Delhi Touchless Palmprint Database. We used them to test the effect of alignmen