Docstoc

A Combined Method for Finger Vein Authentication System

Document Sample
A Combined Method for Finger Vein Authentication System Powered By Docstoc
					                                                              (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                   Vol. 9, No. 7, July 2011

  A Combined Method for Finger Vein Authentication
                      System
                 Azadeh Noori Hoshyar                                                Assoc. Prof. Dr. Ir.Riza Sulaiman
              Department of Computer Science                                        Department of Industrial Computing
              University Kebangsaan Malaysia                                         University Kebangsaan Malaysia
                    Bangi, Malaysia                                                          Bangi, Malaysia
                a_noori_h@yahoo.com                                                         rs@ftsm.ukm.my


                                                      Afsaneh Noori Hoshyar
                                                 Department of Industrial Computing
                                                  University Kebangsaan Malaysia
                                                          Bangi, Malaysia
                                                       Afsa1986@yahoo.com



Abstract— Finger vein as a new biometric is developing in
security purposes. Since the vein patterns are unique between
each individual and located inside the body, forgery is extremely            Vein patterns are located inside the body. Therefore, it
difficult. Therefore, the finger vein authentication systems have         provides a high level of accuracy due to the uniqueness and
received extensive attention in public security and information           complexity of vein patterns of the finger. It is difficult to forge.
security domains. According to the importance of these systems,           Epidermis status cannot effect on recognition system [2].
the different techniques have been proposed to each stages of the         Finger vein systems provide user-friendly environment.
system. The stages include image acquisition, preprocessing,              Therefore, finger vein is a good candidate for authentication
segmentation and feature extraction, matching and recognition.            and security purposes.
While the segmentation techniques often appear feasible in
theory, deciding about the accuracy in a system seems important.              According to the importance of finger vein authentication
Therefore, this paper release the conceptual explanation of finger        system, this paper proposes a system as shown in figure1.
vein authentication system by combining two different techniques
in segmentation stage to evaluate the quality of the system. Also,
it applies Neural Network for authentication stage. The result of
this evaluation is 95% in training and 93% in testing.

   Keywords- Finger Vein authentication; Vein recognition;
Verification; Feature extraction; segmentation

                       I.    INTRODUCTION
    A wide variety of systems require the reliable personal                          Figure 1. The scheme of finger vein authentication system
authentication schemes to confirm or identify an individual                    In the proposed system, different filters are applies for pre-
requesting their services. The purpose of these schemes is                processing stage. Since there are different techniques on
ensuring that only a legal user and no one else can access to             segmentation stage of authentication systems such as matched
provider services. Among different authentication traits such as          filter [3], morphological methods [4], repeated line tracking
fingerprints, hand geometry, vein, facial, voice, iris and                method [5] and maximum curvature points in image profiles
signature, finger vein authentication is a new biometric                  [6], the lack of experiment on combining two different
identification technology using the fact that different person            techniques of “gradient-based threshold” and “maximum
has a different finger vein patterns. The idea using vein patterns        curvature points in image profile” was found to improve the
as a form of biometric technology was first proposed in 1992,             quality of verification system, while the previous studies
while researches only paid attentions to vein authentication in           considered just a single technique for segmentation purpose. In
last ten years. Vein patterns are sufficiently different across           next step, Neural Network is applied to evaluate the quality of
individuals, and they are stable unaffected by ageing and no              training and testing, finally Neural Network is trained and
significant changed in adults by observing. It is believed that           tested for pattern recognition purpose.
the patterns of blood vein are unique to every individual, even
among twins [1].




                                                                     15                                  http://sites.google.com/site/ijcsis/
                                                                                                         ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                  Vol. 9, No. 7, July 2011
    Experimental results of this work show that the system is
valid for user authentication purpose even in high security                                                Input image
environments, as it was the initial intention given the nature of
human finger vein.

     II.   FINGER VEIN AUTHENTICATION SYSTEM                                                                 cropping

The steps of finger vein authentication system are explained in
the following.
                                                                                                          Reducing noise
A. Image Acquisition
   The first step in finger vein authentication system is
capturing the image of finger veins. The quality of captured
image helps to identify the veins of fingers as well. Image                                              Increasing contrast
Acquisition can be done in two ways; i) using infrared-
sensitive digital camera with wavelength between 700nm to                                     Figure 3. Image enhancement process
1000nm and banks of LEDs ; ii) using digital camera with
CCD sensor and IR filter which is located on the camera with             C. Segmentation and Feauture extractions
wavelength 700nm to 1000nm and banks of LEDs .                               In this stage, the enhanced finger vein image is segmented
    Therefore, as shown in figure 2, the Near-infrared rays              and the features are extracted. Since there are different methods
generated from a bank of LEDs (light emitting diodes)                    for segmentation, this paper propose the combination of two
penetrate the finger and are absorbed by the hemoglobin in the           segmentation methods as "Gradient-based thresholding using
blood. The areas in which the rays are absorbed (veins) thus             morphological operation" and "Maximum Curvature Points in
appear as dark areas in an image taken by a CCD camera                   Image Profiles" to segment and extract the features. The
(charge-coupled device) located on the opposite side of the              features of first segmentation method are merged with features
finger. The CCD camera image will be transferred to PC for               of second segmentation method to obtaine an accurate record
next step of authentication [7].                                         for each finger vein images.
                                                                               1) Gradient-based thresholding using morphological
                                                                            operation: In this segmentation method, the gradient of
                                                                            image by alpha filter is created. Then, thresholding is
                                                                            performed on gradient of image. The high gradient values
                                                                            which are more than threshold value in the image fall as
                                                                            edge (vein). After the vein determination in an image, the
                                                                            morphological operations are employed to make an image
                                                                            smoother. The proposed morphological operations are
                                                                            „majority‟ to remove extra pixels, „openning‟ to smooths
                  Figure 2. Image Acquisition system[8]                     the contour of image and breaks narrow passages, „bridge‟
                                                                            to connects the neighbor pixels which are disconnected.
    As stated above, the better quality can make recognition                The original and obtained image after first segmentation
system more accurate. For this purpose, the noise is reduced on             method (includes performing gradient, thresholding,
next step.                                                                  morphological operation) are shown in figure 4.
B. Pre-Processing
    As the image has been taken by camera has redundant parts
which needs to be cropped. Therefore, only the central part of
finger vein image can be taken in Matlab by a simple line;

    I2= imcrop (I, rect);                                   (1)

   Where „I‟ is an image and „rect‟ is the position for
cropping.
                                                                                          a                             b
    The next step in this section is reducing the noise of finger
                                                                            Figure   4.   a) Original     image    b)   Obtained   image   after   first
vein image to improve segmentation. Since the captured image                              segmentation
has much noise, therefore it needs to be improved for getting
better quality. For this purpose, the enhancement Functions                  The total process for the "Gradient-based thresholding
such as ‟medfilt2‟, „medfilt2‟ can be employed. As the final             using morphological operation" method is shown as figure 5.
step for image pre-processing, the image contrast can be
increased using commands in Matlab such as „histeq‟. Figure 3
shows the total process for enhancing the image.




                                                                    16                                    http://sites.google.com/site/ijcsis/
                                                                                                          ISSN 1947-5500
                                                                            (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                                 Vol. 9, No. 7, July 2011
                               Gradient of image

                                                                                        The total process for the "Maximum Curvature Points in
                                 Thresholding                                        Image Profiles” method is shown as figure 5.
                                                                                                                     Calculating the cross
                                                                                                                   sectional profile of image
                               'majority' operation


                                                                                                                     calculating curvatures
                                 'open' operation



                                                                                                                     Calculating the Local
                                'bridge' operation                                                                  maximum of curvatures




                               Segmented image                                                                     Calculating the scores of
                                                                                                                        center points

        Figure 5.Total process of first method for finger vein extraction
                                                                                                                   Assigning scores to center
    2) Maximum Curvature Points in Image Profiles: In this                                                                   points
segmentation method, the curvatures of image profiles are
checks, then, the centerlines of veins are obtained by                                      Figure 5. Total process of second method for finger vein extraction
considering the positions where the curvatures of a cross-                               The following features in the first and second methods of
sectional profile are locally maximal. The centerlines are                           "Gradient-based thresholding using morphological operation"
connected to each other; finally, the vein pattern is achieved.                      and "Maximum Curvature Points in Image Profiles" are
This method is robust against temporal fluctuations in vein                          extracted to train Neural Network.
width and brightness (N.Miura, A.Nagasaka, and T.Miyatake
2005).                                                                               The extracted features for the first method are as follows.
The algorithm for achieving the pattern can be divided into 2                            Sum(~BW2(:)) : The number of black pixels in the
stages;                                                                                   segmented vein image.
 Extracting the centreline positions of veins: The first step of                        Bwperim(BW2): Perimeter of foreground(veins) in
  algorithm is to detect the centerline positions. For this                               segmented vein image.
  purpose, the cross-sectional profile of finger vein image is
  calculated to obtain the intensity value of each pixel along the                       Bwdist(BW2): Number to each pixel that is the distance
  line in an image. In created matrix of intensity, when the                              between the pixel and the nearest nonzero pixel of BW2.
  intensity is positive, it is considered as curvature until it                          Bwarea(BW2): The area of the foreground (veins) in
  becomes negative again. The maximum differences of                                      segmented vein image.
  intensities between two pixels are considered as a vein pixel
  in a row of matrix.
 Connecting center positions of veins: For connecting the                           The extracted features for the second method are as follows.
  center positions, all the pixels are checked. If a pixel and two                       Cross sectional profile of segmented vein image in
  neighbors in both sides have large values, the horizontal line                          vertical direction: Sum of the intensities of pixels in
  is drawn. If a pixel and two neighbors in both sides have                               segmented vein image in vertical direction.
  small values, a line is drawn with a gap at a pixel position.
  Therefore, the value of a pixel should be increased to connect                         Cross sectional profile of segmented vein image in
  the line. The last condition on connecting the center positions                         horizental direction: Sum of the intensities of pixels in
  of veins is a pixel has large value and two neighbors in both                           segmented vein image in horizental direction.
  sides have small values, a dot of noise is created in pixel
  position, and therefore the value of a pixel should be reduced.                        Cross sectional profile of segmented vein image in
  Figure 6 shows the result of second segmentation.                                       oblique1 direction: Sum of the intensities of pixels in
                                                                                          segmented vein image in oblique1 direction.
                                                                                         Cross sectional profile of segmented vein image in
                                                                                          oblique2 direction: Sum of the intensities of pixels in
                                                                                          segmented vein image in oblique2 direction.
                                                                                         Curvatures score: Sum of the calculated scores of
                                                                                          curvatures in segmented vein image.

                 a                                    b
                                                                                     D. Matching and Recognition by Neural Network
                                                                                        The table which is created using the combination of
    Figure 6. a) Original image b) Obtained image after second segmentation
                                                                                     "Gradient-based thresholding using morphological operation"



                                                                                17                                    http://sites.google.com/site/ijcsis/
                                                                                                                      ISSN 1947-5500
                                                                          (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                               Vol. 9, No. 7, July 2011
and "Maximum Curvature Points in Image Profiles" methods is
applied for training Neural Network and also estimating the
quality of training and testing for proposed model. This assess
is done by comparing the true output and the output of the
model.
    For training Neural Network the table is divided into two
tables of training and testing. Therefore, training table has data
are used for training purpose and testing table has data are used
for testing purpose. Also another two tables are created as
training output and testing output. The data of training and
testing output considered as the name of image. Therefore, the
Neural Network has been trained and then simulated to assess
the model quality by comparing the true output and the model
output.
    In training Neural Network, the epochs and goal were
considered „200000‟ and „0‟. The best run occurred when the
performance become close to the goal. As figure 6, the
performance becomes „0.183054‟ from „200000‟ which is close
to the goal „0‟.


                                                                                                     Figure 8. The VAF index for training and testing


                                                                                       After training, the Neural Network is simulated using the
                                                                                       simulation command in Matlab as the following.
                                                                                       R=sim(net,Features)
                                                                                       R= The result of Network simulation
                                                                                       Net = Created Neural Network
                                                                                       Features = obtained features from previous section

                           Figure 6. Training process                                  After simulation the R is obtained as Figure 9.
    The result of this trainig are shown as figure 7. It shows the
differences between output and actual output in training and
testing.The blue line is output and the red line is actual output.




                            a                      b
   Figure 7 a) Output and actual output in training b) Output and actual output
              in testing

    The Variance Accounted For (VAF) index which use to
assess the quality of the model is estimated as 95% for training
and 92% for testing as shown in figure 8 .


                                                                                                             Figure 9. Simulation result

                                                                                           R=7.1756‟ shows the image belongs to the 7th person in the
                                                                                       table which was trained in Neural Network. Therefore, R
                                                                                       recognizes the person who is dealing with a system. This




                                                                                  18                                 http://sites.google.com/site/ijcsis/
                                                                                                                     ISSN 1947-5500
                                                                     (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                          Vol. 9, No. 7, July 2011
recognition can be employed in different applications of                               Conference on Bioinformatics and Biomedical Engineering, ICBBE
security.                                                                              Shanghai
                                                                                 [3]   A.D.Hoover, V.Kouznetsova, and M.Goldbaum. 2000. Locating blood
                                                                                       vessels in retinal images by piecewise threshold probing of a matched
                        III.    CONCLUSION                                             filter response. IEEE Transactions on Medical Imaging 19: 203 - 210.
    This paper proposed a combined method for finger vein                        [4]   Walter, T., J. Klein, P. Massin, and F. Zana. 2000. Automatic
                                                                                       Segmentation and Registration of Retinal Fluorescein Angiographies—
authentication system. “Gradient-based threshold” and                                  Application to Diabetic Retinopathy. In InternationalWorkshop on
“Maximum curvature points in image profiles” were combined                             Computer Assisted Fundus Image Analysis: Denmark.
to obtain precious features. The Neural Network was trained by                   [5]   Miura, N., A. Nagasaka, and T. Miyatake. 2004. Feature Extraction of
the features to evaluate the quality of the system. Also, Neural                       Finger- Vein Patterns Based on Repeated Line Tracking and Its
Network was applied to individual recognition.                                         Application to Personal Identification. Machine Vision and Applications
                                                                                       15:194-203.
    Experimental results of this work show that the proposed                     [6]   Miura, N., A. Nagasaka, and T. Miyatake. 2007. Extraction of Finger-
method is valid for user authentication purpose even in high                           Vein Patterns Using Maximum Curvature Points in Image Profiles. In
security environments, as it was the initial intention given the                       IEICE - Transactions on Information and Systems. Japan: Oxford
nature of human finger vein. Results show that the performance                         University Press.
of the system is 95% in training and 93% in testing.                             [7]   Hitachi . 2006. Finger Vein Authentication: White Paper, available from:
                                                                                       http://www.hitachi-
                                                                                       america.us/supportingdocs/services/smart_solutions/Finger_Vein_Authe
                               REFERENCES                                              ntication_White_Paper.pdf
[1]   Yin, P.Y., ed. 2008. Pattern Recognition Techniques, Technology and        [8]   Lin, D. 1997. Computer-Access Authentication with Neural Network
      Applications. Vienna, Austria.                                                   Based Keystroke Identity Verification. In International Conference on
                                                                                       Neural Networks.
[2]   Lian, Z., Z. Rui, and Y. Chengbo. 2008. Study on the Identity
      Authentication System on Finger Vein. In The 2nd International




                                                                            19                                    http://sites.google.com/site/ijcsis/
                                                                                                                  ISSN 1947-5500

				
DOCUMENT INFO
Description: Journal of Computer Science and Information Security (IJCSIS ISSN 1947-5500) is an open access, international, peer-reviewed, scholarly journal with a focused aim of promoting and publishing original high quality research dealing with theoretical and scientific aspects in all disciplines of Computing and Information Security. The journal is published monthly, and articles are accepted for review on a continual basis. Papers that can provide both theoretical analysis, along with carefully designed computational experiments, are particularly welcome. IJCSIS editorial board consists of several internationally recognized experts and guest editors. Wide circulation is assured because libraries and individuals, worldwide, subscribe and reference to IJCSIS. The Journal has grown rapidly to its currently level of over 1,100 articles published and indexed; with distribution to librarians, universities, research centers, researchers in computing, and computer scientists. Other field coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. (See monthly Call for Papers) Since 2009, IJCSIS is published using an open access publication model, meaning that all interested readers will be able to freely access the journal online without the need for a subscription. We wish to make IJCSIS a first-tier journal in Computer science field, with strong impact factor. On behalf of the Editorial Board and the IJCSIS members, we would like to express our gratitude to all authors and reviewers for their sustained support. The acceptance rate for this issue is 32%. I am confident that the readers of this journal will explore new avenues of research and academic excellence.