Document Sample
					   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856

                                              Vikram Singh1 and Kalpna Kashyap2

                                           Department of Computer Science and Engineering,
                                Noida Institute of Engineering and Technology, Ghaziabad (U.P.), India.

                                            Department of Computer Science and Engineering,
                                Krishna Institute of Engineering & Technology, Ghaziabad (U.P.), India.

Abstract: This paper describes a fingerprint identification
technique which is fusion of two approaches: a minutiae           2. METHODS OF FINGER PRINT MATCHING
based and wavelet statistical features. A minutia matching is
widely used for fingerprint recognition but shows poor               2.1 Minutiae Based Method
performance when image quality is degraded. However, its          For comparison of the fingerprint most of the automatic
performance can be enhanced by the combination of other           systems used minutiae matching. Minutiae are the local
technique like wavelet transform method. Wavelets are good        irregularity in the fingerprint pattern. More than 150
for local analysis, so used for extracting local features from    different minutiae have been identified. But generally
fingerprint. The final matching score is estimated using          ridge ending and ridge bifurcation minutiae are used for
fusing matching scores of minutiae based and wavelet based        fingerprint identification [3]. Many known algorithms
method. By measuring False reject rate (FRR) and false            have been developed for minutiae extraction based on
accept rate (FAR) the performance of hybrid fingerprint           orientation and gradients of the orientation fields of the
identification method is calculated.This method is better than
                                                                  ridges [4, 5].
conventional minutiae method for real time verification.
Keywords: Fingerprint, minutiae, wavelet, FRR, FAR.

Fingerprint recognition has been widely adopted for user
identification due to its reliable performance, usability,
and low cost compared with other biometrics such as
signature, iris, face, and gait recognition [1]. It is used in
a wide range of forensic and commercial applications,
e.g., criminal investigation, e-commerce, and electronic                        (a)                          (b)
personal ID cards [2]. Among many current biometric
technologies, fingerprint identification is the oldest and         Figure 1. (a) Different minutiae types, (b) Ridge ending
most popular method used in different commercial and                                      & Bifurcation
security application. Different methods exist for                 Finger prints are the impression formed by the friction
fingerprint identification but minutiae based approach is         ridges of the skin and thumbs. Finger print identification
commonly accepted approach among them. But it shows               has been used for long time because of their immutability
poor performance, when image quality is not good and it           and individuality. Immutability refers to the unalterable
is not accepted in real time authentication. However the          and undying pattern on each finger while individuality
performance of minutiae based approach can be enhanced            refers to the inimitability across individuals. In fact,
by using statistical images obtained from other image             fingerprints are the furrow and ridge patterns on the tip of
processing techniques. This paper examines hybrid finger          fingers. Due to their individuality and immutability
print matching system with the fusion of features of              fingerprints are used for human identification. The
minutiae and wavelet statistical features and end scores          probability of alike fingerprint is about 1 in 1.9x1015. The
are evaluated by using both minutiae based and wavelet            features extracted from any finger print are divided into
based method and the performance of hybrid finger print           two main classes (i) Global or high level features (ii)
identification can be calculated by measuring its False           Local or low level features. Global features are core and
reject Rate (FRR) and False Accept Rate (FAR) .This               delta and Local features are ridge ending and bifurcations
hybrid method is better than conventional minutiae based          (Fig. 3). In general local features are knows as minutiae.
method for real time authentication systems for large data        Figure 2 shows some fingerprint patterns with the core
bases.                                                            point are marked. Many singularity [6] points (edges)

Volume 1, Issue 4 November - December 2012                                                                          Page 44
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856

detection algorithms were investigated to locate core              Fig. 4 Flowchart of Minutiae Extraction Algorithm
points, among them the famous “Pointcore” index method
[7, 8] and the one described in [9]. Edge detection              2.2 Wavelet Transform Based Method Definition
algorithms typically are followed by linking and other        (Wavelet)
boundary detection procedures designed to assemble edge       A wavelet [14, 15, 16] is a waveform of effectively
pixels into meaningful boundaries [10].                       limited duration that has an average value of zero. In
                                                              mathematical term wavelets are mathematical functions
                                                              that cut up data into different frequency components, and
                                                              then study each component with a resolution matched to
                                                              its scale [17]. Fig.5 illustrates comparison wavelets with
                                                              sine waves. Sinusoids are predictable and smooth and
                                                              extend from minus to plus infinity and do not have
                                                              limited duration where as wavelets may be asymmetric
     (a)                (b)              (c)            (d)   and irregular. By Fourier analysis, if signal is splits into
                                                              sine waves it involves various frequencies. While analysis
 Figure 2. Core points on different fingerprint patterns.     of wavelets includes splitting of signals into shifted and
 (a) Tented arch, (b) right loop, (c) left loop, (d) whorl    scaled versions of original (or mother) wavelet.

                                                                     Fig. 5 Comparison of sine wave and wavelet
                                                              A wavelet means a small wave (the sinusoids used in
Figure3. Global and Local features: core, delta and ridge     Fourier analysis are big waves) and in brief, a wavelet is
                 ending, bifurcations                         an oscillation that decays quickly.
                                                              Equivalent mathematical conditions for wavelet are:
Detection of minutiae [11, 12, 13] in finger print image is              2
described by a list of attributes that involves position of
minutiae, direction of minutiae and types of minutiae           | (t) | dt  ;
(ridge ending and bifurcations). Thus finger print pattern                           ............................(1)
depiction consist features of all identified minutiae. When
minutiae set is represented as a point pattern, the                     2
verification problem of fingerprint can be reduced to a
minutiae point pattern matching problem. A trusty               | (t ) | dt  ;
algorithm is used for extraction of features like ridge                            ..............................(2)
endings and bifurcations. The whole idea for extraction of
minutiae mainly includes three components, orientation        Wavelet Transform
field estimation, ridge extraction, and minutiae extraction   Jean Morlet [18] in 1982, introduced the idea of the
and post processing. Figure 4 shows different steps           wavelet transform and provided a new mathematical tool
involves in finger print using minutiae extraction            for seismic wave analysis. Morlet first considered
algorithm.                                                    wavelets as a family of functions constructed from
                                                              translations and dilations of a single function called the
                                                              "mother wavelet" ψ(t). They are defined by:

                                                              The parameter a is the scaling parameter or scale, and it
                                                              measures the degree of compression. The parameter b is
                                                              the translation parameter which determines the time
                                                              location of the wavelet. If |a| < 1, then the wavelet in (3)
                                                              is the compressed version (smaller support in time-
                                                              domain) of the mother wavelet and corresponds mainly to
                                                              higher frequencies. On the other hand, when |a| > 1, then
                                                              ψa,b(t) has a larger time-width than ψ (t) and corresponds

Volume 1, Issue 4 November - December 2012                                                                               Page 45
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856

to lower frequencies. Thus, wavelets have time-widths
adapted to their frequencies. This is the main reason for
                                                               [ai,{d 1i, d 2 i, d 3i}; i  1, 2,...., k ]
the success of the Morlet wavelets in signal processing        Where ‘ai’ is a low resolution approximation of the
and time-frequency signal analysis.                            original image, and dni are the wavelet sub images
Wavelet transform (wt) depicts image as a summation of         containing the image details at different scales (2k) and
wavelets on different levels of resolution. Strength of WT     orientations (n) [20, 21].
is that it presents high temporal localization for high
frequencies whereas for low frequencies attempts good          In the procedure of matching, a template fingerprint
frequency resolution. Hence, WT is a good means for            having wavelet feature are matched against statistical
extracting local features of image and these features are      feature of input (query) stored in feature library using
required for minutiae extracting of finger print image.        Euclidian distance vector formula given below in Eq.(5).
Fingerprint matching using wavelet transform [19] is           Between points X(x1, x2, x3,..., xn) and Y(y1, y2, y3,……..,
shown below in figure 6. Main steps are as follows:            yn) the Euclidian distance computed as follows as shown
I) Wavelet Statistical Feature Extraction.                     below in Eq.(5):
II) Wavelet Statistical Feature Matching
                                                                ( x1  y1)2  ( x2  y2 )2  ( x3  y3 )2  .....  ( xn  yn )2
                                                               Where x(x1......xn) stands for the features of input test
                                                               finger whereas y(y1......yn) stands for the feature of nth
                                                               template finger print in the library. Feature of the
                                                               fingerprint are stored in feature library and whole finger
                                                               print are matched with the input finger print. Then
                                                               minimum of all distances is found. Every input image has
                                                               minimum distances corresponds to wave let matching
                                                               score and this matching score is again used for estimating
  Fig 6: Fingerprint matching using wavelet transforms         the final matching score [22].

                                                               3. THE HYBRID FINGERPRINT MATCHING
                                                               This method consist the synthesis of two discrete sets of
                                                               fingerprint information: minutiae features and wavelet
                                                               statistical features. The hybrid fingerprint matching
                                                               system is shown below in figure 9.
                                                               The matching procedure continues in a following way
                                                               when query image is present.
                                                                    (i) For producing minutiae matching score: the query
                                                                              and templates minutiae features are matched
                                                                              or generating features matching score.
                                                                    (ii) For generating a single matching score minutiae
                                                                              and the wavelet features are united.
      Fig 7: Wavelet Statistical Feature Extraction            In order to further explore the fusion of minutiae based
                                                               and wavelet based fingerprint recognition methods in a
Figure 8 shows the wavelet statistical feature set
                                                               broad sense: the rank, decision and score level fusion are
extraction process:

            Fig. 8 Wavelet Feature Extractions

The decomposition of two dimensional (2D) wavelet on k
octaves of a discrete image b0 [i,j] represents the image in
terms of 3k+1 sub images as in Eq. (4)

Volume 1, Issue 4 November - December 2012                                                                                         Page 46
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856

       Fig. 9 Hybrid Fingerprint Matching System               common range [0, 1] using three points of reference: i) ѓ:
                                                               a point of the region between a genuine score distribution
3.1 Rank Level Fusion                                          and imposter score distribution, ii) α1 and α2 points that
Rank level fusion is a combination of identification ranks     correspond to the extent of the overlap between the
acquired from multiple unimodel biometrics. The rank of        genuine and the imposter distributions, representing their
the registered users in the data base can be the output of     inferior and superior limits respectively.
the matching module of a biometric system. The objective       3. Tanh – estimators: this method for normalization uses
of the rank level fusion is to consolidate the rank obtained   the standard deviation and mean of the genuine score
individually by the methods in order to derivate a             distribution.
consensual rank for each user. Fig.10 [22] shows an            The matching scores generated by comparing the
example of multimodal biometric system of fingerprint          minutiae sets and wavelet statistical features, are
and iris using rank level fusion.                              combined in order to generate a single matching score.
                                                               This is done by sum rule [24, 25, 26, 27].
                                                               4. PERFORMANCE PARAMETERS
                                                               4.1 False Acceptance Rate (FAR):
                                                               FAR [28, 29] is the probability that an unauthorized
                                                               person is incorrectly accepted as authorized person using
                                                               Match Count (MC) and product of Number of Fingers
                                                               (NF) with Total Images per Finger (IF).

                                                                                 FAR% = MC        *100
  Fig. 10 Multimodal biometric system using rank level                                  (NF*IF)
                      fusion approach                          4.2 False Rejection Rate (FRR):
Ross et al. [23] describe three methods to combine the         It is the probability that the system does not detect an
ranks assigned by different matchers. Those are the            authorized person using Miss Match Count (MMC) and
highest rank method, the Borda count method, and the           Number of Fingers (NF).
logistic regression method.
1. Highest Rank: In the highest rank method, the fused                          FRR% = MMC         *100
rank of a user is computed as the lowest (minimum) rank                                (NF)
generated by different matchers.
2. Borda Count: This method uses the sum of the ranks
assigned by individual matchers to calculate the final         5. CONCLUSION
rank.                                                          The hybrid method tries to formulate the best of on
3. Logistic Regression: The performance of the different       minutiae features and wavelet statistical features.
biometric is not uniform. For example a biometric iris         Minutiae offer rich information for fingerprint matching.
image is performing better than hand geometry or face.         But due to problems like skin dryness, wet fingers,
Therefore assigning the corresponding biometric weights        different images of the same finger, low contact pressure
to the rank of individual matchers have been mostly used       and due to degradation in images like ridges lines are not
modifies the borda count method. The weights are               strictly continuous and parallel lines are not well
calculated during the training phase using logistic            separated due to noise, minutiae extraction becomes hard
regression method.                                             and computationally intensive task. This approach gives
3.2 Decision Level Fusion                                      poor performance if image quality is degraded. Some
Decisions of individual biometric classifiers are fused to     enhancement is necessary in order to improve this
compute a combined decision. This level of fusion is also      algorithm. It requires additional information for
known as abstract level fusion because it is used when         matching. Wavelet shows important features for complex
there is access to only decisions from individual              matching of images like texture oriented patterns. This
classifiers.                                                   method combines minutiae and statistical features.
3.3 Score Level Fusion                                         Features are directly extracted from gray scale fingerprint
Following methods are used for score normalization and         image. This hybrid method is suitable for authentication
used to make possible the score level fusion:                  in real time with number of identities enrolled in them.
1. Min-Max: This normalization method is mostly used
when the bound values (minimum and maximum score               REFERENCES
values) are known. The realization of this method will
                                                                 [1] D. Maltoni, D. Maio, A. K. Jain, and S. Prabhakar,
transform all the scores in a common range ([0, 1]).
                                                                   Handbook of Fingerprint
2. Double Sigmoid Function: This normalization
                                                                 Recognition. New York: Springer-Verlag, 2003.
technique consist in transforming all the scores in a

Volume 1, Issue 4 November - December 2012                                                                      Page 47
   International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)
       Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com
Volume 1, Issue 4, November – December 2012                                    ISSN 2278-6856

 [2] Heeseung Choi, Kyoungtaek Choi, and Jaihie Kim,        [19] Shabana Tadvi and Mahesh Kolte, “A hybrid
    “Fingerprint Matching Incorporating Ridge                 system for fingerprint identification,” IJCSE, vol. 2,
 Features with Minutiae”.                                     no. 3, pp.767-771, 2010.
  [3] Lin Hong, “Automatic person identification using      [20] M. Tico, P. Kuosmanen and J. Saarinen, “Wavelet
    fingerprints,” Ph. D. Thesis, 1998.                       domain features for fingerprint recognition,”
 [4] L. Hong, Y. Wan, and A.K. Jain, “Fingerprint             Electronics Letters, vol. 37, no. 1, pp. 21-22, 2001.
    image enhancement: Algorithm and performance            [21] Marius Tico, Erro Immonen, Pauli Ramo, Pauli
    evaluation”, IEEE Trans. Pattern Anal. Machine            Kuosmanen and Jukka Saarinen, “Fingerprint
    Intell, 20 (8), 777– 789,1998.                            recognition using wavelet features,” IEEE
 [5] Seung-Hoon Chae, Jong Ku Kim, Sung Jin Lim,              Proceedings, 0-7803-6685-9/01/$10.0, pp.21-24,
    Sung Bum Pan, Daesung Moon and Yongwha                    2001.
    Chung, “Ridge-based fingerprint verification for        [22] N. Radha, and A. Kavitha, “Rank level fusion
    enhanced security,” International conference on           using fingerprint and iris biometrics,” IJCSE, vol. 2
    consumer electronics, pp. 1-2, 2009.                      no. 6 Dec 2011-Jan 2012, ISSN : 0976-5166.
 [6] “Digital image processing” by Gonzalez, chapters 3     [23] Ross, K. Nandakumar, and A. K. Jain, “Handbook
    - 4.                                                      of Multibiometrics,” New York: Springer-Verlag,
 [7] Q. Zhang and K. Huang, “Fingerprint classification       2006.
    based on extraction and analysis of singularities and   [24] S. Mallat, “A wavelet tour of signal processing,”
    pseudoridges,” 2002.                                      Academic Press, 1998.
 [8]http://www.owlnet.rice.edu/~elec301/Projects00/rosh     [25] Henry Selvaraj, S. Arivazhagan, and L. Ganesan,
    ankg/elec301.htm                                          “Fingerprint verification using wavelet transform,”
 [9] A.Luk, and S.H.Leung, “A two level classifier for        Proceedings of Fifth IEEE International Conference
    fingerprint recognition,” in Proc. IEEE International     on Computational Intelligence and Multimedia
    Symposium on CAS, Singapore, pp. 2625-                    Applications (ICCIMA’ 03), 2003.
    2628,1991.                                              [26] J Kittler, M. Hatef, R. P Duin, and J. G. Matas,
 [10] Maltino, Maio, Jain, and Prabhakar, “A handbook         “On combining classifiers,” IEEE Transactions on
    of fingerprint recognition,” Springer Press.              Pattern Analysis and Machine Intelligence”, vol. 20,
 [11] M. Kucken and A. C. Newell, “Fingerprint                no. 3, pp. 226-239, 1998.
    Formation,” Proceedings of Journal of Theoretical       [27] A Senior, “A combination fingerprint classifier,”
    Biology, pp. 71–83, 2005.                                 IEEE Transactions on Pattern Analysis and Machine
 [12] Wan Azizun Wan Adnan, Lim Tze Siang, and                Intelligence, vol.23, no.10, pp. 1165-1174, 2001.
    Salasiah Hitam, “Fingerprint recognition in wavelet     [28] Shashi Kumar, K B Raja, R K Chhotaray, and
    domain,” Jurnal teknologi, pp 25-42, 2004.                Sabyasachi Pattanaik, “DWT Based Fingerprint
 [13] A. K. Jain, S. Prabhakar and S. Chen, “Combining        Recognition using Non Minutiae Features,” IJCSI,
    multiple matchers for a high security fingerprint         vol. 8, Issue 2, March 2011
    verification system. Pattern Recognition,” Letters.     [29] A. Wahab, S.H.Chin, and E.C.Tan, “Novel
    20(11-13): pp. 1371-1379, 1999.                           approach to automated fingerprint recognition”,
 [14] Rakesh Verma, and Anuj Goel “Wavelet                    IEEE Proc.-Vis. Image Signal Process, vol. 145, no.
    application in fingerprint recognition,” IJSCE,vol.1,     3, pp. 160-166, June 1998.
    issue-4,Sept 2011.
 [15] M. Sifuzzaman, M.R. Islam and M.Z. Ali,
    “Application of wavelet transform and its advantages
    compared to fourier transform,” Journal of Physical
    Sciences, Vol. 13, pp. 121-134, 2009, ISSN: 0972-
 [16] Umair Mateen Khan, Shoab Ahmed Khan, Naveed
    Ejaz and Riaz Ur Rehman, “A fingerprint
    verification system using minutiae and wavelet based
    features,” IEEE International Conference on
    Emerging Technologies, pp. 291-296, 2009.
 [17] Y.Y.Tan “Wavelet theory and its application to
    pattern recognition world,” scientific, 2000.
 [18] Yuliang He, Jie Tian, Xiping Luo, and Tanghui
    Zhang, “Image enhancement and minutiae matching
    in fingerprint verification,” Pattern Recognition
    Letters, vol. 24, pp. 1349-1360, 2003.

Volume 1, Issue 4 November - December 2012                                                                Page 48

Description: International Journal of Emerging Trends & Technology in Computer Science (IJETTCS) Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 1, Issue 4, November – December 2012, ISSN 2278-6856, Impact Factor of IJETTCS for year 2012: 2.524