International Journal of Computer Applications (0975 – 8887)
Volume 98– No.11, July 2014
The Kingdom of Saudi Arabia Vehicle License Plate
Recognition using Learning Vector Quantization
Artificial Neural Network
Yusuf Perwej, Ph.D
Ph.D, M.Tech (Computer
Science & Engg.)
Assistant Professor
Department of Computer
Science & Engg
Al Baha University, Al Baha,
Kingdom of Saudi Arabia
(KSA)
Nikhat Akhtar
Firoj Parwej
M.Tech, B.Tech (Computer
Science & Engg.)
Assistant Professor
Department of Computer
Science & Engg.
Integral University, Lucknow,
India
Assistant Professor
Department of CS & IS
Jazan University, Jazan
Kingdom of Saudi Arabia
(KSA)
ABSTRACT
In the today scenario technological intelligence is a higher
demand after commodity even in traffic-based systems. These
intelligent systems do not only help in traffic monitoring but
also in commuter safety, law enforcement and commercial
applications. The proposed Saudi Arabia Vehicle License
plate recognition system splits into three major parts, firstly
extraction of a license plate region secondly segmentation of
the plate characters and lastly recognition of each character.
This act is quite challenging due to the multiformity of plate
formats and the nonuniform outdoor illumination conditions
during image collection. In this paper recognition of the
license plates is achieved by the implementation of the
Learning Vector Quantization artificial neural network. Their
results are based upon their completeness in the Saudi Arabia
Vehicle License plate character recognition and theirs have
obtained encouraging results from proposed technique.
Keywords
Arabic Character Segmentation, Learning Vector Quantization
Neural Network, Fan-Beam Feature Extraction, Vehicle
License Plate, Extraction.
1. INTRODUCTION
The License Plate Recognition system is a very significant
part of the Intelligent Transportation System [1] which is very
significant for the development in the transport infrastructure
of the world [2]. In the developing countries such as a
Kingdom of Saudi Arabia. The most common solutions to
license plate localization in digital images are through the
implementation of edge extraction in this edge viewpoint is
normally simple and fast, but it is sentient towards noise [3],
histogram analysis in this basic histogram approach is not
competent of dealing with images with cogitable amount of
noise and tilted license plates [4], and morphological
operators in this the localization of license plates via
morphological based approaches is not high strung to noise
but is very slow in execution [5].
Furthermore though there are many prescripts proposed for
License Plate Recognition system. But, there are not single
prescripts can provide good performance in all the
applications in different complicated background such as the
plate is small, Incertitude of edges, low or high illuminated
images, different types of plate, size, dim lighting, character
fonts, syntax,
weather and environment, spacing etc.
Therefore, most of the previous prescripts could not apply for
all the countries in the real world, all the environments, all
types of the Kingdom of Saudi Arabia Vehicle License Plate.
The Kingdom of Saudi Arabia Vehicle license plate
localization and recognition system today will be required to
operate robustly in environments with intricate backgrounds
and light intensity variations. To deal with such problems,
researchers have proposed in this paper uses Artificial Neural
Network techniques [6] for character recognition, using
Learning Vector Quantization Neural Network [7]. Their
results are compared based upon their completeness in the
character recognition. The efficiency of the System can be
further improved by increasing the number of fonts for
training Learning Vector Quantization Artificial Neural
Networks.
2. THE SAUDI ARABIA VEHICLE
PLATE EXTRACTION
The Arabic License plate extraction is the most vital phase in
the Saudi Arabia vehicle license plate recognition system. The
several models have been proposed for extraction and
recognition aside, but very few of them propose both an
extraction as well as recognition. Several of the researchers
shown in research papers rely on the fact that the number
plate edges are appropriately detected and that the ratio of the
width to height of the number plate is a constant [8]. Several
of the researchers shown in research papers try to enforce
morphological operations to highlight the probable regions of
the number plate and afterward filter out all but one. Several
other methods such as color edge detection and fuzzy systems,
edge statistics and morphological operations, and weight
based density maps. In this paper proposed prescript is
designed for the Kingdom of Saudi Arabia Vehicle License
Plate extraction. Input to the system is an image which
contains the license plate, procured from about six meters
away by a digital camera on the front or rear of the vehicle.
The Kingdom of Saudi Arabia vehicle License Plate method
based on Tophat-Bothat transshipment and Line Scanning.
The prescript includes the following major stages [9], which
are RGB to gray-scale sea-change in other words this image is
then converted to a gray scale image the next step is image
pre-processing for which Tophat and Bothat technique is
used, vertical edge detection and image binarization, analysis
and dilation, vertical projection and thresholding, extracting
32
International Journal of Computer Applications (0975 – 8887)
Volume 98– No.11, July 2014
the actual position of the Arabic license plate, filtration and
image enlargement, binarization and smoothing process [10].
These techniques are mended the overall contrast of the image
as shown in the figure 1.
computing and a partition in the vertical direction, and then
Arabic characters segmentation work is ended.
The segmentation the Arabic characters in digital images in
pursuance of to the variable Arabic digit width and regress a
matrix containing the two bounds of each Arabic digit. The
function lay in the outcome only segment whose rectangular
areas are more than min area Arabic digit.
Figure 2. The Saudi Arabia vehicle Character
Segmentation
4. THE SAUDI ARABIA VEHICLE
FEATURE EXTRACTION
Figure 1. The Saudi Arabia vehicle Plate Extraction
3. THE SAUDI ARABIA VEHICLE
CHARACTER SEGMENTATION
The vehicle license plate character segmentation is the
procedure of extracting the characters from the Arabic license
plate image.
The character segmentation is an essential stage because the
extent one can reach in partition of words lines or characters
directly make an impression the recognition rate of the Saudi
Arabia vehicle license plate [11]. The Saudi Arabia vehicle
license plate scanning done from right to left of the license
plate. In this paper, we proposed first row & second row of
Saudi Arabia vehicle license plate character segmentation,
which harmonize pre-processing to disassemble noises,
normalized and then segment Arabic characters & Arabic
numbers based.
The quantization of gray image, a black & white image laid
down by an adaptive threshold, Then, the image is resized in
parameter 100x150 pixels for first row Saudi Arabia vehicle
license plate and 150x100 pixels for second row Saudi Arabia
vehicle license plate, [12] shown in the figure 2. After preprocessing work, we need to split second rows in the second row Saudi Arabia vehicle license plate counterpart by
The transforming the input data into the set of features is
called feature extraction. If the features extracted are
discreetly handpicked it is expected that the features set will
extract the episodic information from the input data in order to
execute the desired task using this diminished representation
instead of the full size input [13]. Feature extraction is a usual
term for prescript of constructing combinations of the
variables to get around these problems while still depicts the
data with substantial precision.
There are numerous causes why feature extraction is a vital
problem in predictive modeling and modern data analysis. In
first cause dimension lack if the problems with a large
number of variables, almost all prediction models suffers from
the curse of dimensionality, some more harshly than others.
Feature extraction can act as an impressive dimension
deficiency agent. We can comprehend the curse of
dimensionality in very intuitive terms when a person is given
too many variables to consider, most of which are
inconsequential or simply non-informative; it is naturally
much harder to make a better verdict. It is therefore desirable
to select a much smaller number of episodic and vital features.
The high dimensional problems also pretence problems for
computation. Infrequently two variables might be uniformly
informative, but are highly correlated with each other this
often causes unhealthy behavior in numerical computation.
Therefore feature extraction is a vital computational
technique. In another reason automatic exploratory data
33
International Journal of Computer Applications (0975 – 8887)
Volume 98– No.11, July 2014
analysis. If numerous classical applications, informative
features are often selected a priori by field experts, i.e.
explorer pick out what they reckon are the vital variables in
making a model. The more often in modern data-mining
applications however, there is an accrescent demand for fully
automated “black-box” type of prognostication models that
are competent of identifying the vital features on their own
[14]. The necessity for such automated systems arises for two
reasons. Firstly, on the one hand, there are the economic
necessities to process large amounts of data in a short amount
of time with meager manual supervision. Secondly, on the
other hand, it is mostly the case that the problem and the data
are so novel that there are simply no field experts who
understand the data well enough to be able to pick out the
vital variables former to the analysis. The beneath such
situation automatic exploratory data analysis becomes the key.
Instead of keeping faith on pre-conceived consideration, there
is a necessity to let the data speak for itself.
Another motive data visualization. If the application of feature
extraction that shares the flavor of exploratory data analysis is
data visualization. The human eye has an amazing ability in
recognizing systematic patterns in the data. At the same time,
we are ordinarily unable to make good sense of data if it is
more than three dimensional. To maximize the use of the ultra
improved human faculty in visual identification, we
oftentimes wish to identify two or three of the most
informative features in the data so that we can plot the data in
a diminished space. The prescript of feature extraction is
described in below [15]. F(x, y) is the severity of the pixel (x,
y), (xc, yc) is the position of the present tracking point of the
line tracking in the image, Rf is the set of pixels inside the
finger’s outline, and Tr is the locus space. Assume that the
pixel in the lower left in the image to be (0, 0), the positive
direction of the x-axis to be rightward in the image, and the
positive direction of the y-axis to be upward inside the image,
and Tr(x, y) to be initialized to 0.
where i and j are rows and columns of the image serially. Fanbeam takes projections at different angles by rotating the
source around the center pixel at θ degree intervals [17].
These projection data are presumed as feature vectors.
It can be seen from the accumulator data of Fan beam that
after 180 degrees the signals encore itself in the head over
heels direction. This is because projections taken from 0 to
180 degrees are exactly equal to the projections taken from
181 to 360 degrees [18]. The average value of the procured
projection data is computed in order to build the feature
vector. For Fan-beam the average of the projections of one
direction which is the average of 55 parallel projections was
computed. The size of feature vector for one numeral is
1×360.
The column of fan beam data of the image is 540 sensor
samples. It is a common observation in Learning Vector
Quantization Artificial neural network system that, if the
network is trained with more number of features the
simulation accuracy increases. The numbers of features are
more in fan-beam feature extraction; Character Recognition
accuracy is superior for fan-beam feature extraction than other
feature extraction technique.
Step 1: In this step determination of the beginning point for
line tracking and the moving-direction attribute
Step 2: In this step ascertains of the direction of the dark line
and agitation of the tracking point
Step 3: In this step updating the number of time points in the
locus space have been tracked
Figure 3. The Fan-beam Geometry
Step 4: In this step encore execution of step 1 to step 3 (N
times)
5. THE PROPOSED MODEL FOR SAUDI
ARABIA VEHICLE LICENCE PLATE
CHARACTER RECOGNITION
Step 5: In this step acquisition of the finger-vein pattern from
the locus Space
In this paper, we are proposing feature extraction techniques
are tried for training and simulating Learning Vector
Quantization Artificial Neural Network using Fan-beam
Transform. Feature Extraction is performed on each
segmented Arabic character [16].
The Fan beam projection is a variation of radon transform.
The fan-beam function computes projections of an image
matrix along specified directions except that the projections
are taken in a dissimilar way from that of radon transform.
Features were computed using fan-beam geometry. For Fanbeam, 55 diverging beams are taken. The first step is to lay
down the distance D from the fan-beam source to the center of
rotation in figure 3. D must be big sufficient to ensure that the
fan-beam source is outside of the image rotation at all angles.
D is taken a few pixels larger than half the diagonal image
distance, where the diagonal image distance is, d= √i2-j2
In this section, a new algorithm for Arabic character
recognition is developed which is based on learning vector
quantization artificial neural network. It is expected that this
procedure works well even in a Kingdom of Saudi Arabia
vehicle license plate low resolution images. In the literature
we can find several vehicle license plate character recognition
methods. In firstly some vehicle license plate matching
viewpoint use cross correlation to all possible template shifts
over the image while others use hausdorff distance or mean
squared error as ours. In secondly some methods used for
number plate recognition were optical character recognition
and formula based recognition. As learning vector
quantization artificial neural network is an intelligent engine,
it ensures greater accuracy rate along with the better
recognition speed of Kingdom of Saudi Arabia vehicle license
plate.
Artificial Neural Network (ANN) is an interdisciplinary study
of biology and computer science [19], which has been widely
34
International Journal of Computer Applications (0975 – 8887)
Volume 98– No.11, July 2014
used in signal processing, pattern recognition, computer
vision, intelligent control, nonlinear optimization, and so on.
Learning vector quantization artificial neural network
algorithm [20] is one of the most influential methods of
Artificial Neural Network. . Learning Vector Quantization
(LVQ) has been introduced by T. Kohonen as a simple [21], a
universal and efficient classification algorithm and has since
found many applications and extensions.
After extracting features from the Kingdom of Saudi Arabia
vehicle license plate Arabic character images, it’s time to
construct the network. In this paper, the number of input
nodes is the dimensions of image’s eigenvector, and the
number of output nodes is the binary encoding with the
number of Saudi Arabia vehicle license plate character
(Arabic character 28, Arabic numeral character 10).
A Learning Vector Quantization network has a first
competitive layer and a second linear layer. The competitive
layer learns to classify input vectors in much the same way as
the competitive layers. The linear layer transforms the
competitive layer's classes into target classifications defined
by the user. We refer to the classes learned by the competitive
layer as subclasses and the classes of the linear layer as target
classes [22].
The core of Learning Vector Quantization neural network is
based on the nearest-neighbor method of calculating the
Euclidean distance. The distances between each input vector
and competitive layer neural nodes can be calculated [23], and
the output node which is of minimum gap is designated as a
winning node.
d(X,Wc) = min{d(X,Wi)}, (i = 1, 2, · · · , n), where X is the
input vector, Wi is the reference vector,
d(X,Wi) is the gap between X and Wi, and Wc is the winner
subclass. The following equations define the basic Learning
Vector Quantization algorithm process
When i = c, if X and Wc belong to the same class
Wc(t + 1) = Wc(t) + α (t)[X(t) −Wc(t)],
if X and Wc belong to the different classes
Wc(t + 1) = Wc(t) − α (t)[X(t) −Wc(t)],
When i 6= c
Wi(t + 1) = Wi(t),
Where 0 < α (t) < 1, and learning rate α (t) is ordinarily made
to decrease monotonicity with time. It plays a very vital role
in network convergence.
Figure 4. Schematic depiction of LVQ neural network.
Both the competitive and linear layers have one neuron per
(sub or target) class. Thus, the competitive layer can learn up
to S1 subclasses. These, in turn, are combined with the linear
layer to form S2 target classes. (S1 is always larger than the
S2.)
For example, assume that the neurons 1, 2, and 3 in the
competitive layer all learn subclasses of the input space that
belongs to the linear layer target class no. 2. Then competitive
neurons 1, 2, and 3, will have LW2,1 weights of 1.0 to neuron
n2 in the linear layer, and weights of 0 to all other linear
neurons. Thus, the linear neuron output a 1 if any of the three
competitive neurons (1, 2, and 3) win the competition and
output a 1. This is how the subclasses of the competitive layer
are combined into target classes in the linear layer.
In short, a 1 in the ith row of a1 (the rest to the elements of a1
will be zero) effectively select the ith column of LW2,1 as the
network output. Each such column contains a single 1,
corresponding to a conspicuous class. Accordingly, subclass
1s from layer 1 get put into different classes, by the LW2,1a1
multiplication in layer 2.
We know ahead of time what fraction of the layer 1 neurons
should be classified into the various class outputs of layer 2,
we can specify the elements of LW2,1 at the start. We have to
go through a training procedure to get the first layer to
produce the correct subclass output for each vector of the
training set.
By the iterative learning, the input vector X will be assigned
to the class which the reference vector W belongs to. The
class of each input vector can be procured through the
competitive learning process.
35
International Journal of Computer Applications (0975 – 8887)
Volume 98– No.11, July 2014
Figure 6. The Saudi Arabia Vehicle License Plate
Segmentation
Figure 5. The LVQ network architecture.
After extracting features from the Kingdom of Saudi Arabia
vehicle license plate Arabic character images, it’s time to
construct the network. In this paper, the number of input
nodes is the dimensions of image’s eigenvector, and the
number of output nodes is the binary encoding with the
number of Saudi Arabia vehicle license plate character
(Arabic character 28, Arabic numeral character 10). The
learning vector quantization artificial neural network
combines competitive learning with supervision [24]. Target
vector is in log sigmoid form (identity matrix). The learning
rate is 0.01 for training the network.
6. EXPRIMENTAL RESULTS OF THE
KINDOM OF SAUDI ARABIA VEHICLE
LICENCE PLATE RECOGNITION
We tested Kingdom of Saudi Arabia Vehicle License Plate
images which received from the actual system, these vehicle
images are very different background, such as Ignis, license
angles, size and type, Sandstorm, and dimensions from
camera to vehicles, colors, light conditions in the Kingdom of
Saudi Arabia environment, these vehicle images are RGB
true-color image. In the license plate location module, most
of limitations in the previous methods (restricted in case:
uncertainty of edges, various types, low or high ignis images,
colors, sun light & night light) were solved by our proposed
method by using learning vector quantization artificial neural
network.
We implemented experiment with PC Intel(R) Core(TM)2
Duo CPU T7250 @2.40GHz, RAM 4.00 GB, Windows 7 32bit Operating System and MATLAB Version 7.8.0.347
(R2009a).
we have used, subtract operation on the grayscale image to
receive a better image with new severity values that satisfied
for the Arabic image binarization, this is the reason why our
proposed method is very efficient for complex Arabic
background images, night & day images and different Ignis.
Figure 7. The Saudi Arabia Vehicle License Plate
Recognition
we have selected the best values in the fixing the Saudi Arabia
vehicle license plate region step, this value contented for
Saudi Arabia vehicle license plate dimensions. We can see
that, the precision rates of our method are very high such as
Table 1. In this paper, we have proposed an improved method
to segment characters & numbers in both first row and second
row of Saudi Arabia vehicle license plate. In the characters &
numbers recognition module, we have used learning vector
quantization artificial neural network to train for characters &
numbers with noises, so the computing time and precision
improved. In the using network, we have used the image
processing technology for pre-processing to receive high
quality of characters & numbers before putting in the trained
networks, so precision rates of the system are higher in the
Table 1. In this paper graphical data are shows in Saudi
Arabia Vehicle License Plate Segmentation shows in figure 6
and Saudi Arabia Vehicle
36
International Journal of Computer Applications (0975 – 8887)
Volume 98– No.11, July 2014
Figure 8. The Overall Kingdom of Saudi Arabia Vehicle
License Plate System
Table 1. Results of Character Recognition. Test results of
LP Location task, LP Segmentation task, LP Recognition
task
quantization artificial neural network algorithm to recognize
characters & numbers of the Kingdom of Saudi Arabia
vehicle license plate, these networks used independently for
Arabic characters & numbers. The learning vector
quantization artificial neural network has trained with noises
in the training task. Arabic Character & number images
processed by the pre-processing task, which receive high
quality of character & number images for the using network
task to improve precision of the system. It is observed that, as
fan beam feature extraction method has more features for
training the learning vector quantization artificial neural
network thus its simulation precision is higher. Using Fanbeam for feature extraction learning vector quantization
artificial neural network is trained. We have tested our
improved vehicle license plate recognition system in the
Kingdom of Saudi Arabia vehicle license plate images taken
from actual system with different conditions like lightening
conditions (night & day), Sandstorm, license angles,
illumination, size and type, colors and reflected light. Our
vehicle license plate recognition method is more effective
than some existing methods earlier, the efficiency of
computing time & accuracy is improved and very satisfied for
all types of Kingdom of Saudi Arabia vehicle license plate
and Kingdom of Saudi Arabia environment and also more
number of Arabic fonts can be used for training the network
for improving the precision of the Arabic character
recognition.
8. REFERENCES
[1] P. Comelli, P. Ferragina,M. N. Granieri, and F. Stabile,
“Optical recognition of motor vehicle license plates,”
IEEE Trans. Veh. Technol., vol. 44, no. 4, pp. 790–799,
Nov. 1995.
[2]S.-L. Chang, L.-S. Chen, Y.-C. Chung, and S.-W. Chen,
“Automatic license plate recognition,” IEEE Trans.
Intell. Transp. Syst., vol. 5, no. 1, pp. 42–53, Mar. 2004.
[3] V. Abolghasemi, A. Ahmadyfard, An edge-based coloraided method for license plate detection. Image and
Vision Computing 27, pp. 1134-1142, 2009.
[4]Dubey, P., “Heuristic Approach for License Plate
Detection”, IEEE Conference on Advanced Video and
Signal Based Surveillance, 15-16 Sept. 2005, pp. 366370.
[5]W. Pan, R. An, Morphology and auto -correlation based
method of fast locating vehicle license plate. In: Proc. Of
the 2nd Internat. Conf. Advanced Computer Control,
vol.3, pp. 116-119, 2010.
[6] B.D Ripley, Pattern Recognition and Neural Networks,
Cambridge University Press ISBN 0 521 46986 7, 1996.
[7] A. Sato, K. Yamada, Generalized learning vector
quantization, Advances in NIPS, MIT Press, pp. 423429, Vol. 7, 1995.
License Plate Recognition figure 7 finally the Overall
Kingdom of Saudi Arabia Vehicle License Plate System
shows in figure 8.
7. CONCLUSION
In this paper, we proposed an improved vehicle license plate
recognition system for all types of the Kingdom of Saudi
Arabia vehicle license plate, which included of characters
segmentation and character recognition. In the characters
recognition module, we have used a learning vector
[8]C. Anagnostopoulos, I. Anagnostopoulos, E. Kayafas, and
V. Loumos. “A License Plate Recognition System for
Intelligent Transportation System Applications”, IEEE
Trans. Intell. Transp. Syst., 7(3), pp. 377– 392, Sep.
2006.
[9]M. Wu, L. Wei, H. Shih and C. C. Ho. “License Plate
Detection Based on 2-Level 2D Haar Wavelet Transform
and Edge Density Verification”, IEEE International
Symposium on Industrial Electronics (ISlE), pp: 16991705, 2009.
37
International Journal of Computer Applications (0975 – 8887)
Volume 98– No.11, July 2014
[10]T.-H. Wang, F.-C. Ni, K.-T. Li, and Y.-P. Chen, “Robust
license plate recognition based on dynamic projection
warping,” in Proc. IEEE Int. Conf. Netw., Sens. Control,
2004, pp. 784–788.
[20] A. Sato, K. Yamada, Generalized learning vector
quantization, Advances in NIPS, MIT Press, pp. 423429, Vol. 7, 1995.
[11] Dr. Firoj Parwej , “ A Perceptive Method for Arabic
Word Segmentation using Bounding Boxes by
Morphological Dilation”,
for published in the
International Journal of Computer Applications (IJCA)
USA , Volume 71, No.1, Pages 1 – 7, ISSN 0975 –
8887, June 2013.
[22] K.L. Wu and M.S Yang, Alternative learning vector
quantization, Pattern Recognition Journal, pp. 351-362,
Vol. 39, 2006.
[12] Youngwoo Yoon, Kyu-Dae Ban, Hosub Yoon, and
Jaehong Kim, “Blob Extraction based Character
Segmentation Method for Automatic License Plate
Recognition System”,
Proceedings of the IEEE
International Conference on Systems, Man and
Cybernetics, Anchorage, Alaska, USA, IEEE ISBN 9781-4577-0652-3, October 9-12, 2011.
[13] Hossan, M.A. “A Novel Approach for MFCC Feature
Extraction”, 4th International Conference on Signal
Processing and Communication Systems (ICSPCS), pp.
1-5, Dec 2010
[14]P.D. Gader, M. Khabou, Automatic feature generation for
handwritten digit recognition. IEEE Trans. Pat- tern
Analysis and Machine Intelligence 18(12) 1256{1262
(1996)
[15] D. Trier , A. K. Jain and T. Taxt, “Feature extraction
methods for character recognition-A survey”, Pattern
Recognition, vol. 29, no. 4, (1996), pp. 641-662.
[16]G. L. Zeng "Nonuniform noise propagation by using the
ramp filter in fan-beam computed tomography", IEEE
Trans. Med. Imag., vol. 23, no. 6, pp.690 -695 2004.
[17]F. Noo, M. Defrise, R. Clackdoyle, and H. Kudo, "Image
reconstruction from fan-beam projections on less than a
short scan", Phys. Med. Biol., vol. 47, pp.2525 -2546
2002.
[18] J. W. Beattie, “Tomographic Reconstruction from Fan
Beam Geometry Using Radon's Integration Method”,
IEEE Transactions on Nuclear Science, vol. 22, no. 1,
(1975) February, pp. 359-363.
[19] Yusuf Perwej , Firoj Parwej, “A Neuroplasticity (Brain
Plasticity) Approach to Use in Artificial Neural
Network” for published in the International Journal of
Scientific & Engineering Research (IJSER), France ,
Vol.3, Issue 6, pp 1- 9, ISSN 2229 – 5518, June 2012.
IJCATM : www.ijcaonline.org
[21] T. Kohonen, The Self-organizing Map, Proceedings of
the IEEE, pp. 1464-1480, 1990.
[23] M. Biehl, A. Ghosh, B. Hammer, Dynamics and
Generalization Ability of LVQ Algorithms, The Journal
of Machine Learning Research, pp. 323-360, Vol. 8,
2007.
[24]C. Zhan, X. Lu, M. Hou, and X. Zhou, “A LVQ-based
neural network anti-spam email approach”, ACM
SIGOPS Operating Systems Review Volume 39 , Issue,
USA, 2005 pp. 34-39.
9. AUTHOR’S PROFILE
Dr. Yusuf Perwej Assistant Professor in the Department of
Computer Science & Engineering Al Baha University, Al
Baha , Kingdom of Saudi Arabia (KSA). He has authored a
number of different journal and paper. His research interests
include Soft Computing, Artificial Neural Network, Machine
Learning, Pattern Matching, Pattern Recognition, Artificial
Intelligence, Image Processing, Fuzzy Logic, Genetic
Algorithm, Robotics, Bluetooth and Network. He is a member
of IEEE.
Nikhat Akhtar Assistant Professor in the Department of
Computer Science & Engineering Integral University,
Lucknow, India. She has authored a number of different
journal and paper. His research interests include Soft
Computing, Swarm Intelligence, Storage Technology,
Artificial Neural Network, Cryptography, Pattern Matching,
Pattern Recognition, Artificial Intelligence, Network Security,
Fuzzy Logic, Network and Database. He is a member of
IEEE.
Firoj Parwej Assistant Professor in the Department of
Computer Science , Jazan University , Jazan , Kingdom of
Saudi Arabia (KSA). He has authored a number of different
journal and paper. His research interests include Soft
Computing, Artificial Neural Network, Machine Learning,
Pattern Matching & Recognition, Artificial Intelligence,
Image Processing, Fuzzy Logic, Network and Database. He is
a member of IEEE.
38