WAVELET PERFORMANCE IN BIOMETRIC IDENTIFICATION
SYSTEM ACCORDING TO USERS INCREASE
Juan José Fuertes
1
, Carlos Manuel Travieso
2
and Jesús B. Alonso
2
1
Instituto Interuniversitario de Investigación en Bioingeniería y Tecnología Orientada al Ser Humano (I3BH)
Universitat Politècnica de València, I3BH/LabHuman, Camino de Vera s/n, 46022, Valencia, Spain
2
Signals and Communications Department. Institute for Technological Development and Innovation in Communication
Universidad de Las Palmas de Gran Canaria (ULPGC), Campus de Tafira, Edificio de Telecomunicación
E-35017, Las Palmas de Gran Canaria, Spain
Keywords: Wavelet, Biometrics, Hand identification System, Palmprint Texture, Pattern Recognition.
Abstract: This work shows a simple and robust biometric identification system through the use of the palmprint. It
proves the efficiency of the wavelet transform regardless of users’ number. Firstly, the hand palm image
with scale, rotation and translation invariance is isolated from the hand image recorded. Then, the “wavelet
transform” is used to extract the texture features from gray-scale images. Three wavelet families, haar,
daubechies and biortogonal are studied in order to get the best recognition rate. 1440 hand images of 144
people with 10 samples each one have been acquired by means of a commercial scanner with 150 dpi
resolution. Support Vector Machine (SVM) is the main classifier used as identifier in closed mode. A
recognition rate of 99.83% for 50 users and 99.76% for 144 users demonstrate the strong performance of
wavelet transform in biometrics according to users increase.
1 INTRODUCTION
In the competitive business world today, the need and
demand for a biometric physical security solution has
never been higher. Common biometric techniques include
fingerprints, hand or palm geometry, retina, iris, or facial
characteristics. Behavioural character includes signature,
voice (which also has a physical component), keystroke
pattern, and gait among others. Nowadays, most of the
security systems developed into the society are based on
hand image analysis (Masood et al., 2008; Pavesic et al.,
2004), especially in the texture of the hand palm since
they provide people with higher security in relation to
authentication systems (Zhang, 2004) and offer a good
balance of performance characteristics while they are
relatively easy to use. Palmprint analysis offers many
advantages similar to the other technologies such as small
data collection, resistant to attempt to fool a system, ease
of use and difficult technology to emulate a fake hand. In
this context, wavelet analysis plays an important role in
biometric systems: the wavelet filter allows the users to
extract the main features of their hands and to be
differentiated between them (Zhang et al., 2007).
There are however several challenges to beat. Besides
high proprietary hardware costs and size, the aging of the
hands of individuals, the lack of accuracy of the
technology and the biometrics inability to not recognize a
fake hand pose a challenge. To overcome these
drawbacks, Liu et al., 2007, showed a research about the
use of wavelet transform in palm-print. Classifying with
the ISODATA algorithm got a 95% of identification
accuracy with 180 palm-print from 80 people. Masood et
al., 2008, developed a palm-print system using wavelet
transforms. 50 people took part in the session (10 samples
per person), reaching a 97.12% of accuracy with the
combination of different wavelet families. Before, Goh et
al., 2006, had presented a palm-print system made up of
75 individuals. It was based on wavelet transform and
Gabor filter, with a verification result of 96.7% and an
Equal Error Rate (EER) close to 4%. Other authors have
also proposed different techniques of palm-print analysis:
Guo et al., 2009, described a BOCV system, (Binary
Orientation Co-Occurrence Vector) based on the linking
of six Gabor features vectors. 7752 samples from 193
people were taken. The error rate was 0.0189%. Zhang et
al., 2009, presented a novel 2D+3D palm-print biometric
system made up to 108 individuals. The EER was
0.0022%. Badrinath and Gupta, 2009, proposed a
prototype of robust biometric system for verification
which uses features extracted using Speeded Up Robust
Features (SURF) operator of human hand. The system was
tested on IITK database and PolyU database. It had FAR =
0.02%, FRR = 0.01% and an accuracy of 99.98% at
original size. The system addressed the robustness in the
context of scale, rotation and occlusion of palm-print.
482
Fuertes J., Travieso C. and B. Alonso J..
WAVELET PERFORMANCE IN BIOMETRIC IDENTIFICATION SYSTEM ACCORDING TO USERS INCREASE.
DOI: 10.5220/0003770604820487
In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (MPBS-2012), pages 482-487
ISBN: 978-989-8425-89-8
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
Zhang et al., 2010, presented an online multispectral
palmprint system with the requirements of real-time
applications. A data acquisition device is designed to
capture the palmprint images under Blue, Green, Red, and
near-infrared (NIR) illuminations in less than 1 s. The
results show that the Red channel achieves the best result,
whereas the Blue and Green channels have comparable
performance but are slightly inferior to the NIR channel.
Recently, Yue et al., 2011, proposed an algorithm to speed
up the identification process, where the intrinsic
characteristics of the templates of each subject are used to
build a tree, and then, perform fast nearest neighbor
searching with assistance of the tree structure. The
proposed two strategies are 30%-50% faster than brute
force searching. Li et al., 2011, showed a simple efficient
scheme for 3-D palmprint recognition. After calculating
and enhancing the mean-curvature image of the 3-D
palmprint data, line and orientation features are extracted.
Then, the features are fused at either score level or
feature level, reaching a recognition rate of 99.79%.
This paper focuses on the use of 2-D Discrete Wavelet
Transform (DWT) in order to get a simple and robust
biometric identification system using the texture of the
hand palm. Firstly, the hand palm image processing with
scale, rotation and translation invariance (ROI) is
obtained. Then, it will discussed and analyzed several
wavelet biometric families providing its advantages and
disadvantages, showing identification and verification
results, and concluding with the biometrics of the future.
The general block diagram of the system is shown in
Figure 1. After getting the ROI and applying different
filter levels, the recognition rate will be reached with
Support Vector Machine (SVM) (Steinwart & Christmann,
2008), a supervised classifier used to authenticate people.
Figure 1: Functional block diagram of the system.
Figure 2: (a) Database hand image, (b) Fingers detection
of the hand.
This paper is set up as follows: section 2 introduces the
image processing necessary to extract the palm features
presented in section 3; section 4 shows the classification
system used in this work and section 5 explains the
experiments performed; finally, a brief conclusion is given
in section 6 together with the future work.
2 IMAGE PROCESSING
The 1440 hand-images belonging to the 144 users are
acquired thanks to a 150 dots per inch general scanner,
and they are stored with 256 gray levels, 8 bits per pixel
(Figure 2 (a)). The size of these images is set to
1403x1021 pixels after scaling them by a factor of 20% to
facilitate later computation (see Table 1). When a hand is
detected, it is pre-processed to extract the hand-contour
and then, the region of interest (ROI) of the palm. It let us
analyze the palm-print texture and consequently the
identity of the people.
Table 1: Properties of database hand images.
Properties of the hand images contained in the database
Size
80% original size
Resolution
150 dpi
Colour
256 grey levels
File size
1405 Kbytes
Data matrix dimension
14031021
Firstly, we convert each image from 256 gray levels to
a binary image through and adaptive threshold obtained
empirically with training samples. To work out the
biometric features, we localize the 4 fingers of the hand
through 8 initial points of the Figure 2 (b) to detect the
tops and the valleys of the fingers.
Figure 3: Palmprint extraction from hand images:
detection of the valleys.
SVM
classifier
WAVELET PERFORMANCE IN BIOMETRIC IDENTIFICATION SYSTEM ACCORDING TO USERS INCREASE
483
Finding out the maximum of the contour between the
2 points of each finger the ends are obtained. Finding out
the minimum between the 2 consecutive points of different
fingers the valleys are obtained. At this time, the ROI is
extracted after lining up the valley of the little finger with
the valley of the hearth-index finger (Figure 3), in order to
not depend with the hand position. It has a vertical size of
300 pixels and the horizontal size can vary some pixels
depending on the fingers gap.
3 FEATURE EXTRACTION:
DISCRETE WAVELET
TRANSFORM (DWT)
Next step is to process the hand palm image with the
purpose of emphasizing its discriminative characteristics.
The Discrete Wavelet Transform (Villegas & Pinto, 2006)
filters the image separating the thin details from the thick
details of the palm-print image. In this work, 3 wavelet
families have been compared: ‘haar’ or ‘db1’,
‘daubechies5’ and ‘biortogonal5.5’ (see Figure 4 for
different wavelet waveforms).
Figure 4: Several wavelet families: haar, daubechies,
coiflet and symmlet.
We have used a successive chain of low filters with
cutoff=π/2. It lets to emphasize the difference between the
diverse gray levels. The size of the palm-print images is
reduced to a different set of values after applying the
successive filters. The DWT of a signal f(t) has the form
of (1), where ψ(t) is a family of wavelet functions
(Gonzalez and Wood, 2008):
Taking one user palmprint approximation A
k-1
, we can
obtain the k-th level factorization applying a successive
chain of filters, resulting in the approximation (A
k
),
horizontal (H
k
), vertical (V
k
) and diagonal (D
k
) detail
images (Figure 5). In this work, the original palmprint has
been split in approximation and diagonal images, studying
the response of the first one due to its high recognition
rate.
Figure 5: Block diagram of the filter to split the palmprint
into the four images.
After applying filter levels, the image is reduced to
several image sizes to facilitate people recognition. In the
results section, the system accuracy will be shown
according to image size, emphasizing the best wavelet
level.
4 CLASSIFICATION SYSTEM:
SUPPORT VECTOR MACHINE
A general supervised identification/verification system is
divided into two fundamental blocks: training and test, just
as it is shown in the Figure 6. In that figure, the data
capture and the extraction of biometric information are
observed. With these parameters it is modelled a feature
score which is used in the test step.
Figure 6: Block diagram of a general supervised
identification/verification system.
To evaluate the hand features we have used a support
vector machine (SVM) as classifier (Steinwart &
Christmann, 2008). RBF kernel is used in SVM with the
Gaussian function
when it works as
TEST
TRAINING
Obtaining data
Parameters Extraction
Test
(Identification)
Decision
Hand Database
Obtaining data
Parameters Extraction
Score: creating of
biometric feature
Storage
(1)
BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing
484
identifier in closed mode where only known users can
appear in the test data and are taken into account for the
systems performance evaluation. Also the lineal kernel is
studied.
To identify that an input hand belongs to a fixed
identity, we calculate the distance of the input hand
features to the separator hyperplane of the SVM which
models the hand of each identity. The highest distance
belongs to the accepted identity.
It would be possible to say that SVM builds a
hyperplane of separation in the entrance space in two
possible modes: in the first, it is converted the input space
into higher dimension characteristics space, by means of a
(nucleus) non lineal transformation; in the second, it is
built the optimum hyperplane of separation (MMH,
Maximal Margin Hyperplane). This hyperplane maximizes
the distance of the vectors which belong to different
classes. Thus, if S is a set of N vectors
€
where
i=1...N, each vector
belongs to one of the two
identifying classes as y
i
€ {-1,1}. If the two classes are
linearly separable, then a unique optimum hyperplane
defined by equation 2 exists:
· + b = 0 (2)
It provides a greater margin of separation among the
classes, and it divides S leaving all the vectors of the same
class in the same side of the hyperplane. Support Vector
Machines (SVM) are based on a bi-class system, where
only two classes are considered. In particular for this
work, we have worked on identification system, and for
this reason, we have built a one-versus-all strategy for
SVM, where the second class will consist of the rest of the
classes, under closed mode (Bin et al., 2000). To express
the similarity between biometric patterns, in our case
palm-print modality, we have used the recognition rate.
The higher the recognition rate, the better the system
performance.
5 EXPERIMENTS AND RESULTS
In this section we have evaluated the response of three
wavelet families for 50 and 144 users. Four images of
each user are chosen randomly as training samples and the
remaining six images are used to test the system. The
results are shown in average (%) and typical deviation
(std). Each test has been done 10 times using lineal and
RBF SVM kernels, finding the result through a cross-
validation strategy. The supervised classification has been
carried out with SVM_light (Cortes & Vapnik, 1995).
Figure 7 ilustrates the result after applying the 3
rd
level of
wavelet filter.
At this time, the extracted and filtered palmprint is
reduced to different image sizes in order to get a faster
system while the accuracy increases. Then, the image is
introduced to the classifier in order to get the recognition
rate. In figures 8 and 9 the performance rate of the haar
wavelet according to size increase after applying the
second and third filter levels for 144 users are shown.
Figure 7: Left, the original palmprint. Right, palmprint
factorization in the third wavelet level.
When the fourth level was applied to palmprints the
recognition rate decreased considerably up to 84.14% ±
2.80 and 86.33% ± 1.33 with lineal and RBF classifier.
Figure 8: Identification rate for second level applying Haar
wavelet: progress according to palmprint sizes.
Figure 9: Identification rate for third level applying Haar
wavelet: progress according to palmprint sizes.
Once the image is filtered, the higher the palmprint
size the better the system recognition. This happens
because important discriminative features are ruled out
when the image is reduced to a small size. When the
applied filter is daubechies5 or biortogonal5.5, the
recognition rate progress is similar to haar filter. In Tables
2 and 3 the highest recognition rate with lineal and RBF
kernels for 50 and 144 users and the three wavelet families
are shown.
WAVELET PERFORMANCE IN BIOMETRIC IDENTIFICATION SYSTEM ACCORDING TO USERS INCREASE
485
Table 2: Wavelet identification results for 50 users.
IDENTIFIER – 50 USERS
Features
Lineal
Recognition Rate
RBF
Recognition
Rate
Wavelet Haar
3 filters
99.33% ± 0.00
99.17% ± 0.06
Wavelet db5
3 filters
99.66% ± 0.22
99.50% ± 0.06
Wavelet
bior5.5
3 filters
99.83% ± 0.06
99.67% ± 0.22
Table 3: Wavelet identification results for 144 users.
IDENTIFIER – 144 USERS
Features
Lineal
Recognition
Rate
RBF
Recognition
Rate
Wavelet Haar 3
filters
99.54% ±0.01
99.73% ± 0.03
Wavelet db5
3 filters
99.54% ± 0.01
99.31% ± 0.12
Wavelet bior5.5
3 filters
99.76% ± 0.01
99.65% ± 0.02
According to tables 2 and 3, the recognition rate is
similar regardless of database users’ number, unlike other
texture algorithms like derivative method. The best
recognition rate for 144 users (99.76% ± 0.01) is reached
with lineal kernel for biortogonal5.5 wavelet. This result is
similar to 99.83% reached for 50 users. The third level of
the low pass filter provides the suitable value of the
texture to maximize the interclass relationship in order to
get the large discriminative information. This is possible
because it makes better use of the frequency-space
resolution, and consequently the classifier is able to
differentiate the users properly.
We can compare the wavelet performance with other
algorithms applied to the same database (Travieso et. al
2011), as the derivative method, where if the number of
users increases, the recognition rate decreases
considerably (see Table 4).
Table 4: Comparison with other algorithms for 50 and 144
users.
IDENTIFIER – 50 & 144 USERS
Algorithm
Recognition
Rate (50 users)
Recognition Rate
(144 users)
Wavelet
Haar 3
filters
99.83% ± 0.06
99.76% ± 0.01
Derivative
method
99.99% ± 0.01
99.46% ± 0.13
Gabor
filter
99.66% ± 0.01
99.73% ± 0.02
In next section we will discuss about these results and
the system performance. It will be also introduced the
future work and possible improvements in our work.
6 CONCLUSIONS AND FUTURE
WORK
In this paper, a deep study about the performance of
wavelet transform in a biometric identification- system is
shown. A recognition rate of 99.76% is reached using
palmprint features of the users’ hand.
The results depict the reliability of the filters due to
the high rates obtained for the three wavelet families for
50 and 144 users. It makes the growth of the database
possible independently the number of users, reaching
similar recognition levels.
The hand palm texture extraction algorithm is
intuitive, simple and quick, with a computational load
similar to geometrical parameter extraction.
It is important to obtain the palmprint with the method
explained in the section 2, with scale, rotation and
translation invariance. It lets the classifier differentiate the
users. Moreover, the proposed biometric feature is well
adapted to the SVM classifier, which identifies the feature
degree of simplification necessary for the best
performance. This is very important for the system
successful, and it happens for both 50 and 144 users.
A higher accuracy system could be built combining
Wavelet transform with other texture algorithms or with
geometrical methods. Fuertes et al., 2010, and Ferrer et al.,
2007, proposed two studies about the performance of
geometrical and texture methods were shown,
demonstrating the improvement of the system when some
algorithms were merged. It is possible to combine them at
score or decision level, getting a safer and higher accuracy
system respectively.
One drawback we have to mention it is the necessity
of operating with clean hands. Painted or dirty hands
would cause an identification mistake. In this case a
combined geometric/palm biometric feature would be
more advisable.
Many applications and algorithms have been proposed
in the last years, but few of them are really used. Our
future work is focusing on new 2D+3D technologies, and
in the improvement of reliable algorithms which can be
merged with other techniques, as face recognition.
Specifically for this research line, it will be interesting a
deep study about the stability of the wavelet algorithm for
many users. It is not the same to test a system with 100 or
1000 users in order to get a system regardless of number
of users. If the number of users is long and unknown,
wavelet analysis can be an excellent technique to
recognize people.
BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing
486
ACKNOWLEDGEMENTS
This work has been partially supported by "Catedra
Telefónica - ULPGC 2009/10" (Spanish Company), and
partially supported by Spanish Government under funds
from MCINN TEC2009-14123-C04-01.
Authors want to thanks to Processing Digital Signal
Division from Institute for Technological Development
and Innovation in Communications (PDSD-IDETIC) for
the work which all the Division has done on the building
of the Palm-print database.
REFERENCES
Badrinath, G. and Gupta, P. (2009). Robust Biometric
System Using Palmprint for Personal Verification.
Advances in Biometrics, volume 5558 of Lecture
Notes in Computer Science, pages 554-565. Springer
Berlin/Heidelberg.
Bin, Z., Yong, L., and Shao-Wei X. (2000). Support
vector machine and its application in handwritten
numeral recognition. Proceedings of the 15th
International Conference on Pattern Recognition,
volume 2, pages 720-723, Barcelona, Spain.
Cortes, C. and Vapnik, V. (1995). Support Vector
Networks. Machine Learning, 20: 273-297.
Ferrer, M. A., Morales, A., Travieso, C. M., and Alonso,
J.B. (2007). Low Cost Multimodal Biometric
identification System Based on Hand Geometry, Palm
and Finger Print Texture. 41st Annual IEEE
International Carnahan Conference on Security
Technology, pages 52-58, Otawa, Canada.
Fuertes, J. J., Travieso, C. M., Ferrer, M. A., and Alonso,
J.B. (2010). Intra-modal biometric system using hand-
geometry and palmprint texture. Security Technology
(ICCST), 2010 IEEE International Carnahan
Conference on, pages 318-322, San José, California.
Goh, M. K., Connie, T., Teoh, A. B. J., and Ngo, D. C. L.
(2006). A Fast Palm Print Verification System,
Proceedings of the International Conference on
Computer Graphics, Imaging and Visualization, pages
168-172, London, UK.
Gonzalez, R. C. and Wood, R. E. (2008). Digital Image
Processing. Prentice Hall, 3rd edition.
Guo, Z., Zhang, D., Zhang, L., and Zuo, W. (2009).
Palmprint verification using binary orientation co-
occurrence vector. Pattern Recognition Letters,
30:1219-1227.
Li, W., Zhang, D., Zhang, L., Lu, G., and Yan, J. (2011).
3-D Palmprint Recognition with Joint Line and
Orientation Features. Systems, Man, and Cybernetics,
Part C: Applications and Reviews, IEEE Transactions
on, 41(2): 274-279.
Liu, F., Lin, C. X., Cui, P.Y., and Dong, T. (2007).
Palmprint recognition based on ISODATA clustering
algorithm. Proceedings of the 2007 International
Conference on Wavelet Analysis and Pattern
Recognition, volume 3, pages 1129-1133. Beijing,
China.
Masood, H., Mumtaz, M., Butt, M. A. A., Mansoor, A. B.,
and Khan, S. A. (2008). Wavelet Based Palmprint
Authentication System. Proceedings of IEEE
International Symposium of Biometric and Security
Technologies, pages 1-7, Islamabad, Pakistan.
Pavesic, N., Ribaric, S.,and Ribaric, D. (2004). Personal
Authentication Using Hand-Geometry and Palmprint
Features. Proceedings of the Workshop on Biometrics
at ICPR04, Cambridge, UK.
Steinwart, I. and Christmann, A. (2008). Support Vector
Machines. Information Science and Statistics,
Springer.
Travieso, C. M., Fuertes, J. J., and Alonso, J. B. (2011).
Derivative Method for Hand Palm Texture Biometric
Verification. Security Technology (ICCST), 2011
IEEE International Carnahan Conference on,
Barcelona, Spain.
Villegas, O.O.V. and Pinto R.E. (2006). Procesamiento
Digital de Imágenes en el Dominio Wavelet. IEEE
Looking Forward, 13: 13-16.
Yue, F., Zuo, W., Zhang, D., and Li, B. (2011). Fast
palmprint identification with multiple templates per
subject. Pattern Recognition, 32(8): 1108-1118.
Zhang, D.D. (2004). Palmprint Authentication.
International Series on Biometrics, volume 3, pages
256.
Zhang, D., Guo, Z., Lu, G., Zhang, L., and Zuo, W.
(2010). An Online System of Multispectral Palmprint
Verification. IEEE Transactions on Instrumentation
and Measurement, 59(2): 480-490.
Zhang, D., Kanhangad, V., Luo, N., and Kumar, A.
(2009). Robust palm print verification using 2D and
3D features. Pattern Recognition, 43(1): 358-368.
Zhang, L., Guo, Z., Wang, Z., and Zhang, D. (2007).
Palmprint Verification using Complex Wavelet
Transform. IEEE International Conference on Image
Processing, volume 2, pages 417-420, San Antonio,
Texas, USA.
WAVELET PERFORMANCE IN BIOMETRIC IDENTIFICATION SYSTEM ACCORDING TO USERS INCREASE
487
