درخت تصمیم گیری دودویی فازی مبتنی بربهینه سازی کلونی مورچه ها برای تایید سیستم بند انگشت دست

In the recent trends of touch-less biometric authentication systems, hand knuckles from dorsal part of the hand is gaining popularity as a potential candidate for verification/recognition in variety of security applications. However, most of the available knuckle verification systems offer fixed security achieved for desired level of accuracy which cannot meet the varying levels of security requirements. This paper presents a bimodal knuckle verification system which is designed to meet a wide range of applications varying from civilian to high security regions. We use ant colony optimization (ACO) to choose the optimal fusion parameters corresponding to each level of security. The developed verification system utilizes fuzzy binary decision tree (FBDT) which is aimed at decision making in two classes: genuine (accept) and imposter (reject) using matching scores computed from the knuckle database. The FBDT is implemented using fuzzy Gini index for the selection of the tree nodes. The experiments are carried out on four publicly available HongKong PolyU knuckle databases named as: left index, right index, left middle and right middle with four bimodal systems: left–right index, left–right middle, left index–middle and right index–middle. The experimental results from these four bimodal knuckle databases validate the contributions of the proposed work.

Hand based biometrics have been the most acknowledged for personal authentication. Not only due to its superior performance which is required for the high security applications, but also for their high distinctiveness, user convenience and acceptance. The hand biometrics can be broadly divided into two categories: the palmer part and the dorsal part. The palmer part of the hand generally spans the areas close to palm. The widely used biometric attributes extracted from this part of the hand are: fingerprints (Jain et al., 2007, Maltoni et al., 2003 and Ratha and Bolle, 2004), palmprint (Huang et al., 2008, Jia et al., 2008, Kong et al., 2006 and Zhang et al., 2003) and finger/palm veins.1 The dorsal part of the hand occupies the area behind the palmer part and most of the usable biometric modalities from this part are: hand geometry/shape (Jain et al., 1999, Sanchez-Reillo et al., 2000 and Yoruk et al., 2006), hand veins (Kumar and Prathyusha, 2008, Wang and Leedham, 2006 and Wang et al., 2008) and finger knuckles (Kumar and Ravikanth, 2009, Kumar and Zhou, 2009, Li et al., 2004, Nanni and Lumini, 2009, Woodard and Flynn, 2005a and Woodard and Flynn, 2005b). The palmer region of the hand is supposed to have more informational details than dorsal part and several unimodal/multimodal biometric systems have been attempted using fingerprint and palmprint biometrics. However, people leave their palm/handprint unconsciously wherever they touch and which increases the possibilities of imposter attacks on these security systems. The biometric modalities from the dorsal part of the hand are therefore gaining popularity. Owing to the touch-less acquisition, they have less chance of imposter attacks and being a non-active part of the hand there is less possibility of information degradation, especially compare to the palmer part. Though, the biometric traits extracted from the dorsal part have shown to be less informative in compare to the high detailed fingerprint or palmprint. Despite promising effort using hand shape biometrics (Yoruk et al., 2006), the geometry of hand is considered as less distinctive (Jain et al., 1999). Beneath the skin, the vein patterns from the dorsal part of the hand possess high textural details. But, its acquisition in outdoor environment is itself a challenge and requires costly infrared thermal sensors with several complex algorithms for pre-processing (Wang & Leedham, 2006). The current popularity of the knuckle biometrics for personal authentication is due to its high textural details, simple acquisition with ordinary digital camera, and high user convenience without use of any pegs or hand docs (Jain et al., 1999). However, since knuckle from one finger has a small surface area, knuckles from more than one finger have been considered for research purposes. Out of the five fingers of a hand, knuckles from index and middle fingers have largely been investigated in the literature. The knuckles of these two fingers from left and right hands have also been attempted to make a bimodal/multimodal system. This paper aims at developing a bimodal system by investigating several possible combinations of the four knuckle points from both the hands. The presented approach relies on soft computing pattern classification approach using fuzzy binary decision tree (FBDT) for designing a verification system and evolutionary optimization technique utilizing ant colony optimization (ACO) for optimally select the fusion parameters in the knuckle bimodal system.

This paper has several promising outcomes: (I) the potentiality of FBDT based classification is for the first time utilized for the knuckle verification. The use of Gaussian and Trapezoidal membership functions has provided four modes for training of FBDT. Two splitting measures: fuzzy information gain and fuzzy Gini index are considered for computing the tree-nodes and the FBDTs generated from these two splitting criteria is trained with the four modes. However, we found FBDT with Gini index and Gaussian membership in single mode having the most promising results and continued for the bimodal system. (II) The ACO is for the first time employed for the selection of fusion parameters in the knuckle bimodal systems. Four score level fusion strategies are tested in the bimodal system out of which one is selected as optimal one by ACO. Unlike the PSO which needs the sigmoid function for the discrete values of velocity and position updates, the ACO is defined for discrete domain and uses the pheromone based probabilities to decide the direction of the search on the optimal path. Further, the local and the global updates of these probabilities from the previous search paths are incorporated for the optimal selection of fusion parameters in the proposed bimodal system. (III) The knuckle bimodal systems developed in this work is designed to meet varying security requirements and adaptive to various security applications. The different security levels are quantified in terms of CFA and ACO is employed for optimally select the fusion parameters corresponding to each of them. The presented approach to the knuckle bimodal verification system is novel in its applicability to various security applications; as earlier efforts offer fixed security and designed for a chosen level of security. The experimental results obtained from four combinations of PolyU knuckle databases validate the contribution from this paper. However, we found these bimodal systems operating on almost similar performance and selection of a bimodal system is as per user convenience and availability. Out of four fusion rules tested in this work, sum rule has shown consistently chosen by ACO and hence can used in any knuckle bimodal system.