A Direction-Based Vascular Pattern Extraction Algorithm for Hand Vascular Pattern Verification

This paper proposes an improved vascular pattern extraction algorithm for person verification applications. The proposed direction-based vascular pattern extraction (DBVPE) algorithm is based on the directional information of vascular patterns. It applies two different filters to the input images: row vascular pattern extraction filter (RVPEF) for effective extraction of the abscissa vascular patterns and column vascular pattern extraction filter
(CVPEF) for effective extraction of the ordinate vascular patterns. We use the combined output of both filters to
obtain the final hand vascular patterns. Unlike the conventional hand vascular pattern extraction algorithm, the directional extraction approach prevents loss of the vascular pattern connectivity. To validate the DBVPE algorithm, we used a prototype system with a DSP processor. The prototype system shows approximately a
three-times better false acceptance rate (FAR) than the conventional single filter algorithm.

INTRODUCTION
Biometrics is a technology that verifies or identifies persons using their physiological or behavioral characteristics. As our society is moving at light speed into the information age, the demand for biometric technology is growing at a much faster rate.

There are several modalities for biometric person verification: fingerprint systems, retina/iris systems [1], [2], hand geometry systems, hand vascular systems, etc. Each of
these systems has merits and demerits. For example, fingerprint technology has an advantage in its implementation size and uniqueness of biometric features [3], [4], but it has severe problems in usability [5]. Usability is defined as the percentage of an unspecified population that is capable of using a technology. Because of the lack of usability, it is difficult to apply fingerprint technology to work places such as factories, construction sites, and places with inferior environments.

Hand geometry technology [6] shows excellent performance in the usability measure, but it suffers from a relatively high false acceptance rate (FAR) measure. In spite of its disadvantage, it occupies the second largest market share in the US biometrics market, mainly because of its good usability. Hand vascular technology is excellent in the usability measure and has many advantages because it uses biometric features inside the human body rather than on the surface, and this results in a very stable verification performance during long periods of time.

The problem with conventional hand vascular technology [7], [8], however, is that the vascular pattern is extracted without considering the directional characteristics of the vascular patterns. As a result, there is some loss of connectivity of vascular patterns and verification performance degradation in terms of its FAR. The impact of this problem is more severe with subjects who have relatively thin vascular patterns or a contracted vascular pattern due to
various conditions, such as exposure to cold. This paper specifically focuses on minimizing loss of the vascular pattern connectivity.

In order to reduce the impact of the loss of vascular pattern connectivity, we propose a new algorithm, called the direction-based vascular pattern extraction (DBVPE) algorithm. It applies two different preprocessing filters to the input images: row vascular pattern extraction filter (RVPEF) for effective extraction of the abscissa vascular patterns and column vascular pattern extraction filter (CVPEF) for effective extraction of the ordinate vascular patterns. We combine the output of both the filters to obtain the final hand vascular patterns. Our investigation demonstrated that the directional extraction approach substantially reduced loss of vascular pattern connectivity compared to the conventional hand vascular pattern extraction algorithm. The main purpose
of the RVPEF is to effectively extract the abscissa vascular patterns while preserving pattern connectivity and that of the CVPEF is to effectively extract the ordinate vascular patterns while preserving pattern connectivity.

The proposed DBVPE algorithm preserves connectivity information by minimizing loss of both the abscissa and ordinate vascular pattern information. Using DBVPE in constructing our algorithm for hand vascular pattern person verification, we observed substantial success in resolving critical problems of the conventional hand vascular pattern extraction algorithm.

To test and validate the DBVPE algorithm, we devised a prototype system, the hand vascular pattern recognition system
(HVPRS), to implement the DBVPE algorithm. To effectively implement the hardware, we designed the filters to have filter
coefficients to the power of two, which enabled implementation with fixed-point operators only. In this paper, we describe the DBVPE algorithm and the HVPRS in detail
and present the results of the performance evaluation of DBVPE.

To demonstrate the performance improvement of our algorithm over the conventional single filter algorithm, which does not utilize directional information, we first analyzed the conventional algorithm as well as the DBVPE algorithm; this is described in section II. In section III, we then present the implementation details of the DBVPE algorithm and the HVPRS. Section IV contains the results of the performance evaluation and the experimental results. Finally, in section V, we conclude this paper with a presentation of the future direction of the research.

Please visit A Direction-Based Vascular Pattern Extraction to read entire paper.

Sang Kyun Im, Hwan Soo Choi, and Soo-Won Kim

—-------------------------------------------
Manuscript received June 4, 2002; revised Oct. 24, 2002.
This work was supported in part by Biometrics Engineering Research Center.
Sang Kyun Im (phone: +82 2 523 4715, e-mail: isk@tech-sphere.com) is with TechSphere
Co. Ltd., Seoul, Korea.
Hwan Soo Choi (e-mail: hschoi@mju.ac.kr) is with the Department of Information
Engineering, Myong-ji University, Seoul, Korea.
Soo-Won Kim (e-mail: ksw@asic.korea.ac.kr) is with the Department of Electronics
Engineering, Korea University, Seoul, Korea.