Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

3 Paper DSD

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

IETE Journal of Research

ISSN: 0377-2063 (Print) 0974-780X (Online) Journal homepage: https://www.tandfonline.com/loi/tijr20

A Novel Weighted SVM Classifier Based on SCA for


Handwritten Marathi Character Recognition

Surendra P. Ramteke, Ajay A. Gurjar & Dhiraj S. Deshmukh

To cite this article: Surendra P. Ramteke, Ajay A. Gurjar & Dhiraj S. Deshmukh (2019): A Novel
Weighted SVM Classifier Based on SCA for Handwritten Marathi Character Recognition, IETE
Journal of Research, DOI: 10.1080/03772063.2019.1623093

To link to this article: https://doi.org/10.1080/03772063.2019.1623093

Published online: 19 Jun 2019.

Submit your article to this journal

Article views: 1

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tijr20
IETE JOURNAL OF RESEARCH
https://doi.org/10.1080/03772063.2019.1623093

A Novel Weighted SVM Classifier Based on SCA for Handwritten Marathi


Character Recognition
Surendra P. Ramteke1 , Ajay A. Gurjar2 and Dhiraj S. Deshmukh3
1 Department of Electronics & Telecommunication Engineering, Shram Sadhana Bombay Trust College of Engineering and Technology,
Bambhori, Maharashtra, India; 2 Sipna College of Engineering & Technology, Amravati, Maharashtra, India; 3 Mechanical Engineering,
Dr. Babasaheb Ambedkar College of Engineering & Research, Nagpur, Maharashtra, India

ABSTRACT KEYWORDS
The research on handwritten optical character recognition (OCR) of Marathi script is very challeng- Optical character
ing due to the complex structural properties of the script that are not observed in most other scripts. recognition; WOAR-SVM;
This paper gives an OCR framework for handwritten Marathi document classification and recognition Sine cosine algorithm;
system. Due to the large variety of symbols the Marathi characters recognition poses great chal- Modified Pihu method;
Global transformation
lenge and their proximity in appearance. The weighted one-against-rest support vector machines
(WOAR-SVM) assume a noteworthy part to deal with vast feature measures which are utilized for
the classification. Here, a new sine cosine algorithm is proposed for the identification of handwritten
Marathi text. By utilizing different morphological operations the preprocessing is finished and the
Marathi text is flexibly segmented in three levels; line segmentation, word segmentation and charac-
ter segmentation with Modified Pihu method. Various features like statistical, global transformation,
geometrical and topological features are extracted from the preprocessed image by extraction tech-
niques. Result obtained show that various features with WOAR-SVM classifier perform the best by
yielding high accuracy as 95.14%.

1. INTRODUCTION
There are heaps of issues in handwritten character recog-
In digital computer machine the serious research topic is nition when contrasted with documents of machine
simulation of human perusing. Not only the main advan- printed. Since various peoples have distinctive styles in
tages of such exertion were difficult for simulating human composing, pen-tip estimate and in their writing some
reading but also the probability of efficient application peoples have skewness. To overcome this issue every one
in which printed and handwritten character present on of the difficulties makes the researchers to work. In India
document has to be transferred into machine justifi- Devanagari script is an older most one, which is utilized
able format [1]. The automatic character recognition of to write numerous languages like Nepali, Hindi, Marathi,
printed and handwritten written record information has Sindhi and Sanskrit for documentation [9]. Be that as it
an assortment of practical and commercial applications may, the generally preferred language is Marathi; a very
in libraries, banks and post offices [2,3]. In the field of less measure of work has been finished.
image processing, pattern recognition, machine learning
and artificial inelegancy the optical character recogni- In these areas most of the present work is restricted to
tion (OCR) is an examination. OCR is a procedure of English and a few oriental languages. For Indic scripts
converting scanned images of machine printed or hand- the absence of efficient solutions in Marathi language
written text into a computer processable format [4]. All has hampered extraction of information from a histor-
the OCR particularly of records in English language has ical importance and social archives. For text charac-
been comprehensively studied and actualized effectively ter segmentation different techniques have elaborated;
over years [5]. they are wavelet transforms [10,11], curvelet transform
[12,13], and Gradient feature [14,15]. Consequently, for
The OCR comprises two classifications based on data the OCR these techniques are not demonstrated depend-
acquisition process: off-line character recognition and able [16,17].
online character recognition [6–8]. The off-line character
recognition is additionally separated into two sections: In this paper, we exhibit a modified approach for
machine printed and handwritten character recognition. handwritten Marathi text document classification and

© 2019 IETE
2 S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION

recognition process. To begin with, the Marathi text The hybrid LMs are built by joining both neural net-
documents are changed into the image samples. In the works (FNNLMs and RNNLMs) with BLMs. For reason-
wake of preprocessing the noise is expelled and script able correlation with state-of-the-art system and BLMs,
was segmented adaptable in three levels utilizing ‘Mod- assessed in a system with the similar character over-
ified Pihu Method’. After that, various feature extraction segmentation and classification models as previously,
techniques are applied to segmented image to extract different LMs are analyzed utilizing a little text corpus
relevant features from character to form feature vec- used previously. In their work the execution of both
tors. These feature vectors are then utilized by a novel ICDAR-2013 and CASIA-HWDB competition dataset
approach named weighted one-against-rest support vec- is enhanced essentially. The character level accurate
tor machines (WOAR-SVM) classifier for the classifi- rate and correct rate accomplish 95.88% and 95.95% in
cation and to recognize the Marathi word image the CASIA-HWDB, respectively. R Pramanik et al. [21] have
sine cosine algorithm (SCA) optimization algorithm is introduced a novel shape decomposition-based segmen-
used. WOAR-SVM is a binary classification algorithm; tation technique to disintegrate the compound characters
nonetheless, it is utilized to classify the feature database of into prominent shape components. The less number of
the document. Our approach would add to decrease the classes to recognize the shape decomposition lessens the
error, so our proposed strategy achieves high accuracy classification complexity also enhances the recognition
and furthermore, this approach accurately recognizes the accuracy at the same time. At the segmentation area
substantial volume of characters. The rest of the paper the decomposition is done where the two fundamental
is structured as follows: In section 2, the brief review of shapes are joined to form a compound character.
handwritten character identification is provided. Section
3 expounds the proposed method in detail. Section 4 A system for offline recognition cursive Arabic hand-
illustrates experimental analysis and finally, section 5 written text based on hidden Markov models (HMMs)
concludes the paper. is presented by R. Mouhcine et al. [22]. Without explicit
segmentation the system is analytical to perform embed-
ded training. By baseline estimation the statistical fea-
2. LITERATURE SURVEY: A BRIEF REVIEW
tures are extricated and in the word image geometric to
Many of the frameworks have clarified about a variety integrate both peculiarities of the text and pixel distri-
of different strategies for text character segmentation. A bution characteristics. Using HMMs these features are
portion of the works is assessed here. modeled and trained by embedded training. The popu-
lar deep convolutional neural networks (DCNNs) have
Y Zhang et al. [18] have presented a novel adversarial been introduced by C Boufenar et al. [23]. The DCNNs
feature learning (AFL) model to enhance the HCR exe- have adequately replaced the hand-crafted descriptors
cution on restricted data training that incorporates the with network features and appeared to provide preferable
prior knowledge of printed data and writer-independent outcomes than other traditional methods. In machine
semantic features. From accessible handcrafted feature learning, it is one of the quickest developing areas and
methods, the introduced AFL method is distinctive promise to reshape the future of artificial intelligence.
which automatically exploits writer-independent seman- In three different ways the CNN model can be used:
tic features and standard printed data as prior knowledge training the CNN from scratch, from a pretrained model
is learnt objectively. To solve the issues of speed and stor- utilizing transfer learning strategy to leverage features,
age capacity, X Xiao et al. [19] have introduced a global keeping the transfer learning strategy and fine-tune the
supervised low-rank expansion technique and an adap- CNN architecture weights.
tive drop-weight method. A nine-layer CNN intended for
HCCR comprises 3,755 classes and devises an algorithm S Roy et al. [24] have exhibited a novel deep learning
that decreases the computational cost of networks by nine strategy for the recognition of handwritten Bangla iso-
times and in baseline model the network is compressed lated compound character. On the CMATERdb 3.1.3.3
to 1/18 of the original size with 0.21% accuracy drop. dataset a new benchmark of recognition accuracy is
Contrasted with CNN strategy for HCCR, the introduced reported. In different pattern recognition issues the
model is around 30 times speedier and 10 times costlier. greedy layer-wise training of deep neural network has
helped to made critical steps. The authors utilize layer-
The author [20] assesses the impacts of two sorts of wise training to DCNN in a supervised fashion and
character-level NNLMs as FNNLMs (feed forward neural to accomplish faster convergence the training process
network LMs) and RNNLMs (recurrent neural network is augmented with the RM-SProp algorithm. To solve
LMs) for enhancing Chinese handwriting recognition. the problem of recognizing isolated handwritten words
S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION 3

the authors [25] have introduced the use of a new neu- vectors are then utilized by a novel approach based
ral network architecture that combines a deep convolu- on weighted one-against-rest support vector machine
tional neural network with an encoder–decoder called classifier and the SCA optimization algorithm. To opti-
sequence to sequence. The presented architecture distin- mize the WOAR-SVM, the SCA algorithm is utilized to
guishes the contextual and characters with their neigh- select the Marathi text from the handwritten document.
bors to recognize any given word. Under several experi-
ments the author’s models are tested on two handwritten
databases like IAM and RIMES to determine the optimal 3.1 Image Acquisition Phase
parameterization of the model.
Image acquisition is an initial phase of character recogni-
tion system. In this phase the input handwritten or paper
document image is scanned and converted into electronic
3. PROPOSED METHOD FOR MARATHI
form in bitmap images such as JPEG, BMT, TIF and TNG.
CHARACTER RECOGNITION AND
The acquired image is fed to the pre-processing phase
CLASSIFICATION
Figure 2.
The design of the proposed OCR system is shown in
Figure 1. For this character classification and recognition
the preprocessing, segmentation, feature extraction and
optimization techniques are proposed in our method. At
first, the text documents are changed over into the image
samples in the preprocessing stage. Then, the Marathi
script was segmented flexibly in three levels as line seg-
mentation, word segmentation and character segmen-
tation with ‘Modified Pihu Method’ being proposed to
enhance the segmentation accuracy. After segmentation
various features are extracted from the character image.
A typical feature of the handwritten text is the presen-
tation of text created by the author. Feature extraction
stage is to expel the data redundancy. These feature Figure 2: Sample document image

Figure 1: Architecture of the Proposed OCR


4 S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION

The entire image is managed utilizing the global seg-


mentation process and local segmentation dealt with only
sub-images. In the global threshold process, based on the
image intensity values the threshold level is considered.
Here, the global threshold of the image (IM) is f (x, y)
represented as T if the intensity value is less than the
threshold level, then it is set as 0 (black); the rest are set
as 1 (white). The threshold image g (m, n) is given as
follows.

Figure 3: Image after Binarization 1, if f (x, y) ≥ T
g(x, y) = (1)
0, otherwise

3.2 Preprocessing Phase For the segmentation technique the projection pro-
file method is utilized in our method. The segmenta-
The next phase is preprocessing, it manages enhanc- tion technique incorporates the three vital strategies:
ing image quality for better recognition by the system. Line segmentation, Word segmentation and Character
Preprocessing of handwritten document is required to segmentation.
identify and evacuate all undesirable bit patterns which
prompt to lessen the recognition accuracy. The main a) Line Segmentation: In this segmentation method,
objectives of preprocessing are Binarization, noise reduc- initially on every line or row utilizing the horizontal
tion and line removal. After text preprocessing the var- projection profile method the sum of all white pix-
ious feature extraction techniques have been utilized to els is estimated and also the appropriate histogram
extract features for recognition process. of the image is generated as takes after.
• The horizontal histogram of the image is con-
(a) Grayscale conversion: In grayscale conversion, the structed.
stored bitmap images (JPEG, BMT, TIF and TNG) • The distance between proper two histograms is
are changed over to grayscale image format. Here, recognized, based on the threshold value every
in the matrix form the images are available where histogram is separated and saved.
all the values of every element are identical to how • Finally, from the image the segmented line is
bright or dark the pixel at the fitting position should produced.
be colored. b) Word Segmentation: In word segmentation strategy,
(b) Binarization: The binarization process utilizes a to estimate the entirety of every single white pixel
global threshold approach to convert a grayscale the vertical projection profile approach is used. The
image into a binary image. Based on the threshold segmentation of the word is delineated as takes after.
value these procedures increment the processing rate • At first, the vertical histogram for the image is
and diminished the required storage space Figure 3. developed.
(c) Image Noise Removal: In scanning devices the gen- • In every column, discover the number of white
erated noises in image are line segment separated, pixels and by using the histogram the columns
bumps in lines and gaps. The main distortions are with no white pixel are detected.
local variations, dilation and erosion, etc., and fur- • Replace every such column by 1 and change over
thermore it is exceptionally essential to supplant the the unfilled rows as 0 and content words will have
restrictions. The median filtering is utilized to per- unique pixels and save it.
form noise removal. From the image this strategy • From the line the words are segmented based on
decreases the salt and pepper noise. the threshold value and the procedure is rehashed
for each line Figure 4.
c) Character Segmentation using Modified Pihu
3.3 Segmentation Phase
Method: The Modified Pihu method is proposed to
In the segmentation phase the continuous character of overcome the existing method limitations [26,27].
preprocessed image is broken down into sub-images of Figure 5 demonstrates the Marathi word and its
individual character. The segmentation plays a notewor- different components. Amid segmentation the Pihu
thy part in character recognition process. Segmentation method [27] does not evacuate the word header line
process can be sorted as global and local segmentation. and focuses only on the shape of the characters. So,
S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION 5

arrows in Figure 5.

w (Hl )
P
CVWPC = PW (j); for j = 0, 1 ≤ m ≤ Wh (5)
rn, 0

where, the mth row of the first column is rn,0 and Pw (j)
denotes the black pixels and the same procedure is
repeated for the whole word Figure 5.

Figure 4: (a) Segmented Line Image (b) Segmented Word Image Step 4: The IM is segmented using Equation (6)


⎪ true, CVWPC − 95%



⎪P (Hl ) 
w W

⎪ IMj,i ,

j=rn,0 i=1
f (ch_cut) = (6)

⎪true, if (mid(CVWPC − 95%)



⎪ ∪ (5%Pw ∈ lmh − 15%))



Otherwise; false

where, lmh is the lower modifier (LM) height, f (ch_cut)


Figure 5: Character image segmentation using Modified Pihu represents the segmented character and the total charac-
method ter count is L.

the Modified Pihu method is proposed to segment Chw = f (ch_cut)j ; 1 ≤ j ≤ L (7)


the word header line which is delineated in following
advances:
Chw (1) + Chw (2) + · · · + Chw (L)
Chavg_w = (8)
Step 1: Calculate the height (Wh ) and width (Ww ) of the L
segmented word image IMas, During the segmentation of image into subunits, there
may exist a character (CC) due to the presence of LM. The

n
Wh = Y(j); for 1 ≤ j ≤n (2) character width is calculated to check the possibility of
j=1 said cases and compared with Chavg_w (average character
width).

m
Ww = X(i); for 1 ≤ i ≤m (3) Case (i): if Chw > Chavg_w , then the sub-image from CC is
i=1 segmented using the following function
where, Y(j) and X(i) are the number of pixels along the ⎧P (H ) W
w l 
y-axis and the x-axis, respectively, m is the mth pixel along ⎪


⎨ IMj,i , true, if (PB ∈ mid_
x-axis and n is the nth pixel along y-axis. f (CC) =
j=1 i=1
(9)

⎪ area ∪ Chw > Chavg_w )


Step 2: The row with maximum white pixels in the image Otherwise; false
top 30% area is evaluated as
Case (ii): if Chw > Chavg_w ,mid(CVWPC) ≥ 95% then
Pw (Hl ) = max (Hl1 , Hl2 , . . . Hlk ) (4) images are segmented using Equation (6).

W
Here, Hlk = Pw (k, i) represents the header line rows Step 5: Segmentation of modifiers, the upper modifier
i=1 (UM) is segmented using Equation (10).
in the top area of image containing white pixels and the
header line is denoted as Pw (Hl ). ⎧ w
⎨  
W
⎪ h
IMj,i , true, if (Pw ≥ 1)
Step 3: Calculate the CVWPC area, from the bottom left f (UM) = j=PB(Hl ) i=1 (10)


corner up to Pw (Hl )of the word as shown by red color Otherwise; false, iff (Pw = 0)
6 S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION

The LM is segmented using Equation (11). a) Zernike moments: The Zernike moments normal-
⎧ ization aims to influence the recognizing proce-
⎪ (H
l ) 
W

⎪ dure of an object in terms size of image transla-


IMj,i , true,

⎨ j=1 i=1 tion and rotation independent. The Zernike moment
f (lmh) = if (Chh ≥ (Chavg_h + lmh)) (11) with order n and repetition rof a continuous image

⎪ functionf (x, y) is given as

⎪ Otherwise; false,



if (Chh < (Chavg_h + lmh) n + 1 
Znr = f (x, y) [vnr (x, y)] ∗ (12)
3.14 x y
where the height and characters of the LM are lmh and
Chh , respectively. Amid the character recognition pro-
b) Hough Transform: The Hough transform procedure
cess, the advantage is that the modifier optimizes the
is utilized for baseline document detection. It is like-
character class count independently by storing and helps
wise applied to characterize the characters’ parame-
in diminishing the processing time. The separated modi-
ter curves. The Hough transform is given by
fiers are extracted utilizing feature extraction techniques
in the following section. 
L
H(a) = h (xi , yi, a1,... an ) (13)
i=1
3.4 Feature Extraction
c) Fourier Descriptor: For shape analysis the Fourier
The most vital part of the recognition system is feature
transformation is broadly utilized. The transformed
extraction technique. This phase is utilized to evacuate
coefficients are from the shape of the Fourier
data redundancy. The feature extraction can be char-
descriptors to represent the shape in frequency
acterized as extracting the most illustrative data from
domain. The number of coefficients generated from
raw information that limit inside class design variability
the transform is vast, to capture the overall features
while improving. Various feature extraction methods are
of the shape the subsets of coefficient are enough.
characterized in three groups: statistical features, global
The boundary of particular shape has k pixel num-
transformation features and geometrical and topological
bered from 0 to k − 1. Along the contour, kth pixel
features.
has (xk , yk )position. The shape of two parametric
conditions with (x, y) coordinates as s(k) = x(k) +
3.4.1 Statistical Features
i y(k). The discrete Fourier transform of s(k) is
The statistical features are derived from the statistical
points of distribution. They provide low complexity and
1
k−1
−j2π vk
high speed of variation to some extent, also used for b(v) = s(k)e k ; v = 0, 1, . . . k − 1 (14)
reducing the feature set dimension. The following are the k 0
statistical features:
d) Gabor Transform: The variation of the windowed
Fourier transform is a Gabor transform. The win-
a) Zoning: The character frame is divided into several
dow utilized as a part of this case is certifiably not
overlapping or non-overlapping zones. The densi-
a discrete size yet defined by a Gaussian function.
ties of some features in different regions are analyzed
In both spatial and frequency domains the trans-
utilizing zoning approach [28].
form possesses optimal localization properties. The
b) Crossings and Distances: The number of crossing of
2D Gabor transform gives an extracted feature and
a contour by a line segment in a specified direction
it is represented by
is the popular statistical feature. The frame contain-
ing the character is parceled into an arrangement of
regions in different ways and afterward features of x 2 + χ 2 y2 2 xπ
G (χ , φ, η, κ) = exp( ). cos( + ϕ)
each region are extracted. 2 ξ2 κ
(15)
3.4.2 Global Transformation Features
These features are invariant to global deformations like where, x = a cos θ − y sin θ by varying the parame-
rotations and translation. For the purpose of classifi- ters like χ , φ, η, κ the transform can be used better.
cation the continuous signal generally contains more
data that need to be represented. By linear combina- 3.4.3 Geometrical and Topological Features
tion the signal is represented by series of simple well- These features may represent the global and local char-
characterized functions. acter properties and have high resistances to distortions
S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION 7

Figure 6: (a) Shadow features of character, (b) Contour point


detection

and style variations. Some information about the object


contour is required to sort components of the object by Figure 7: Shape of character using features
these topological features.
dk and ak constitute 24 features from 12 sectors, as
a) Shadow Features: The shadow features are com- shown in Figure 7.
puted with rectangular boundary encasing the image e) Extraction of occupancy and end points: The shape
of character is separated into eight octants; the char- profile relies on dk and ak , these features were ini-
acter shadow segment is computed on two perpen- tially proposed with different features such as occu-
dicular sides in each octant, so a total of 24 features pancy and end points of character used in the recog-
are obtained. Figure 6(a) demonstrates the shadow nition system. For deciding occupancy only four
with projection on side length. On scaled image sectors are utilized by tracing the character the end-
these features are processed. points are demarcated. The point is proclaimed as
b) Chain Code Histogram of Character Counter: For a endpoint if no ‘1’ pixel is present otherwise ‘0’and
scaled binary image the contour points of the char- the sector which is additionally noted as shown in
acter image are identified by considering a window Figure 7.
(3 × 3) encompassed by object points as shown in f) Transition feature: Another feature is a transition
Figure 6(b). Chain coding is utilized to represent the feature approach extracted from calculation and
contour, each pixel is allocated as a various code that location, from the back ground to foreground pixels
shows the following pixel direction. In this approach in vertical and horizontal directions. Based on gray
by utilizing the chain coding neighbor contour pixel level sub-images of single characters some extrac-
associated with the the outline coding and points are tion methods work, while others work on connected
captured. symbol segmented from binary raster image and
c) Finding Junctions/Intersection in Character: The skeleton symbols.
junction is suggested as intersection (location) g) View-based features: For correct character recogni-
where the chain code goes more direction in a 8- tion a human usually needs only partial information
connected neighborhood. Thinned and scaled char- about it shape and contour. This feature extraction
acters are divided into segments of pixel size. The method views a set of points that plot in one of
number of open end points and junctions of each four projections of object (left, right, bottom and
segment is calculated. For a character in different top) comprising pixel belonging to character con-
segment the intersection points are unique. tour and have extreme values of one of its coordi-
d) Extraction of distance and angle features: The num- nates. Thus, the characteristics vector is created from
ber of ‘1’ pixel present in k sector is nk with k = the acquired features to portray the given character,
1, 2, . . . , 12. The normalized vector distance (dk ) for which is the base for facilitating optimization and
each sector is the sum of distances of all pixel (‘1’) in classification.
a sector partitioned by the number of pixel present
in that sector, where the coordinates of sector pixel
3.5 SCA for Feature Selection
are (xi , yi ) and the coordinates of center of charac-
ter image are (xm , yn ). This dk is taken as one set of The SCA is the very best method of optimization in such
features and the corresponding angles for each pixel problems. For developing a new optimization algorithm,
for each sector are computed. Another arrangement SCA is utilized. SCA is based on the population opti-
of feature is taken as a normalized angle (ak ). Both mization technique; with a set of random solutions it
8 S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION

starts the optimization process. By an objective function


this random set is evaluated repeatedly and improved by
a set of rules that is kernel for optimization. The SCA
algorithm looks for the optima of problem in optimiza-
tion stochastically and there is no guarantee of finding
a solution in a single run. In the field of stochastic opti-
mization regardless of the difference between algorithms,
the division has two phases: exploration and exploitation
[29]. In the former phase, to find a promising region of
search space with high rate of randomness the optimiza-
tion algorithm combines the random solutions. However,
in random solution there are gradual changes and ran-
dom variation is considerably less than those phases. For
error character minimization problems the fitness solu-
Figure 8: Steps for SCA Algorithm
tion can be simply proportional to the objective function
value.

Initialization is the first step of the feature extracted solu-


tions of the OCR system. The position updating equation
for both phases is proposed as follows.

Yit+1 = Yit + R1 × sin(R2 ) × |R3 pti − Yit | (16)

Yit+1 = Yit + R1 × cos(R2 ) × |R3 pti − Yit | (17)


where, the position of current solution is Yit in i − th
dimension and iteration t − th , R1 /R2 /R3 are random
numbers, the position of destination point is pt and || this
represents the absolute value. The above two equations
Figure 9: Analysis of (a) Recall (b) Precision using proposed
are combined as follows: method

Yit + R1 × sin(R2 ) × |R3 pti − Yit | , R4 < 0.5
Yit+1 =
Yit + R1 × cos(R2 ) × |R3 pti − Yit |, R4 ≥ 0.5
where, the current iteration is T, tmax is the maximum
(18) number of iteration and the constant is A. The pseudo
code of SCA algorithm is presented in Figure 8. Figure 8.
The four main parameters in SCA are R1 , R2 , R3 , and R4 . Steps for SCA Algorithm
The parameter (R1 ) indicates the next regions in space
between destination and solution. R2 defines how far When the iteration counter is higher than tmax the SCA
the movement should be towards or outwards the des- algorithm terminates the optimization process. With the
tination. R3 denotes the random weight for destination above steps the proposed algorithm is theoretically able
and R4 parameter equally switches between the sine to determine the global optimum of problems in opti-
and cosine components. Due to the formation of sine mization due to following reasons:
and cosine this algorithm is called SCA. For finding the
promising regions of search space the algorithm balances • For a given problem SCA creates and improves the
the exploration and exploitation phase also it covers the set of random solutions, so local optima avoidance is
global optimum level. The range of sine and cosine in compared to other single solution-based algorithms.
Equation (16)-(18) is changed in order to balance the two • When sine and cosine functions return values (bet
phases using the following equation. ween −1 and 1) different regions of search space are
explored.
A • Using adaptive range in the functions SCA smoothly
R1 = A − T (19)
tmax transits from exploration to exploitation.
S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION 9

• The best global optimum is stored in variables as Values of χ1+ scales measure the uncertainty dj (yt ) in
destination during optimization and updates their a positive direction, whereas χ1− denotes the same for
positions. negative values of χ1− . The final decision d(yt )was deter-
mined by Equation (23).
The next section employs recognition and classification ⎧
⎨d(yt ) = arg maxc [χj  βi yi K(yi, yt ) + b]
n
which take the optimized output of the feature extraction j
techniques to their input and then determine which class i=1 (23)

it actually belongs to. d(yt ) = arg maxcj [dj (yt )]

χ1− χ2− . . . χj− . . . χc−


3.6 WOAR-SVM-Based Classification and = (24)
χ1+ χ2+ . . . χj+ . . . χc+
Recognition
Collected weights which are used for the WOAR-SVM
The WOAR-SVM is the most extensively used classifier approach in matrix as represented in Equation (24).
for pattern recognition because of its many promising The optimization of is a part of model selection prob-
empirical performance and attractive features. In this lem for WOAR-SVM approach. Using this way the pro-
paper, for the classification the WOAR-SVM classifier is posed method is able to identify the texts. It is observed
implemented which takes the optimized output of the that this identification step helps in improving the OCR
feature extraction techniques to their input and deter- performance by getting clear text. This research also con-
mine which class the character image actually belongs to. siders the handwritten text storage and processing.
WOAR-SVM is a powerful supervised classifier based on
the utilization of weight coefficients. The main goal of it
is error minimization classification and maximization of 4. RESULTS AND DISCUSSION
margin discrimination. In this section, the results of the proposed WOAR-SVM
classifier with SCA have been discussed with real-time
In WOAR-SVM approach the j − th binary prediction of dataset of handwritten and synthetic Marathi characters.
SVM classifier is expressed in Equation (20) for a new From different people the handwritten documentation
observation yt . materials have been composed as proposed databases.
Furthermore, from school, home, and office, the real-

n
time datasets are gathered in this section. The evalua-
dj (yt ) = βi yi K(yi, yt ) + b (20)
tion results of the proposed WOAR-SVM classifier are
i=1
in contrast with those of the current classification, seg-
Generally, the positive sign of dj (yt ) indicated that some mentation and feature extraction strategies are portrayed.
data points yt and the value dj (yt ) represent the certainty Additionally, the analysis of the proposed method is per-
measure in that decision. The proposed WOAR-SVM formed using the FAR (false acceptance rate), accuracy,
classifier has two assumptions: 1) for test sample the vali- precision, recall, F-measure and FRR (false rejection rate)
dation set must have sufficient accurate representation, 2) metrics over the existing algorithms.
for both training and testing samples the validation must
be independent. So the performance of classifier will not 4.1 Dataset Description
verify and results would not be good indicators of the
In this research, a self-created database (handwritten and
performance of classifiers. The linear coefficients χj are
synthetic data) was utilized. Here, for testing and train-
introduced for dj (yt ), the correlation between multiple
ing processes 33 Marathi characters are utilized. All the
binary classifiers is reflected in the final decision.
data samples are divided into two parts after the prepro-
cessing stage: for training purpose 56100 data samples

n
dj (yt ) = χj βi yi K(yi, yt ) + b (21) are reserved while for testing purpose 9900 data samples
i=1 are reserved. In the dataset, for the accuracy and the per-
formance evaluation the training samples are utilized by
where, χj was a function of sign of dj (yt ) as shown in WOAR-SVM classifier.
Equation (22)
 4.2 Evaluation Metrics
χj− for dj (yt ) < 0
χj = (22) For the proposed WOAR-SVM classifier with SCA the
χj+ for dj (yt ) > 0 performance validation metrics are considered for the
10 S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION

Table 1: Performance of evaluation metrics with the pro-


posed method
Script

Synthetic Handwritten
S. No Metrics data data
1 Accuracy (%) 92.80 95.14
2 Precision (%) 93.6 96.5
3 Recall (%) 94.7 98.3
4 FAR (%) 91.24 95.04
5 FRR (%) 90.41 93.52
6 F-measure (%) 94.68 98.1

recognition of optical character in this section. The quan- Figure 10: Analysis of (a) Accuracy (b) F-measure using the pro-
titative metrics used are accuracy, precision, recall, FAR, posed method
FRR and F-measure to evaluate the classification results,
as shown in Table 1.

a) Analysis based on Recall and Precision: In each


analysis the data samples consists of following classes
true positive (T p ), true negative (T n ), false posi-
tive (F p ), false negative (F n ) (Figure 9). By these
four classes the handwritten Marathi characters are
correctly classified. Here, the positive label T p is cor-
rectly identified by the classifier. The negative labels
T n (correctly rejected) were correctly labeled. The
incorrectly classified, negative labels F p and posi-
tive labels F n were incorrectly labeled as negative by
the classifier. In the following equations they are Figure 11: Analysis of (a) FAR (b) FRR using the proposed method
delineated,
Tp c) Analysis based on FAR and FRR: The total number
recall = (25) of the indelicately identified recognition of charac-
Tp + F n
ters is FAR. It is given by

Tp FAR = Fp /(Fp + Tn ) (29)


Precision = (26)
Tp + Fp The FRR is the total number identified the rejected char-
b) Analysis based on Accuracy and F-measure: The acters recognition is given by
weighted mean of precision and recall is F-measure
FRR = Fn /(Fn + Tp ) (30)
which is formulated in Equation (27).
The analysis curve based on the FAR is shown in
Precision ∗ Recall
F - measure = 2 × (27) Figure 11(a). The analysis curve based on the FRR is
Precision + Recall
shown in Figure 11(b). For the reliable recognition sys-
tem the classifier is used normally and the rejection rate
The accuracy comprising of two parameters (FAR and
is reduced for the optimal performance. The false rejec-
FRR) is given by
tion analysis is based on the synthetic and handwritten
Tp + Tn data. While comparing both the handwritten data have
Accuracy = (28) 93.52% FRR and 95.04% FAR.
Tp + Tn + Fp + Fn

Figure 10 shows the analysis of accuracy and F-measure


4.3 Performance Comparison
using the proposed method. While comparing both syn-
thetic and handwritten data the accuracy of the system is In this section, based on classification rate, classifica-
heavily dependent on the variation of the writing styles in tion time and error parameters the performance of pro-
training datasets. For training the accuracy of the system posed method is evaluated. Also the proposed method
improves as we increase the test samples. is compared with different classifiers such as K-NN
S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION 11

Table 2: Performance evaluation with different classifiers Table 4: Error comparison using the proposed method
Classification Classification Classification
S.No Classifier rate (%) time (ms) S.No Algorithm Script Error (%) time (ms)
1 K-NN 88.6 98.04 1 Zheng et al. [30] Synthetic data 6.2 85.37
2 SVM 73.3 43.59 Handwritten data 6.8
3 WOAR-SVM (proposed) 94 42.15 2 Lin et al. [31] Synthetic data 2.5 52.16
Handwritten data 7.9
3 Proposed Synthetic data 1.7 42.15
Handwritten data 0.3

Figure 12: Evaluation of classification parameters using different


classifiers

Table 3: Execution time comparison of the existing method


with modified Pihu method
Existing method Modified Pihu Figure 13: Comparison of error using different algorithm
Image [26] method
Size (pixels) 1169 353 1169 353
Execution time for segmentation (s) 4.76 2.72 Figure 13 illustrates the error comparison between the
proposed method and the existing approaches. The pro-
posed algorithm has less error rates in the range of 1.7%
(K-nearest neighbor), SVM (support vector machine) as to 0.3% for Marathi script recognition than Devanagari
shown in Table 2. script, i.e. these texts are classified more accurately. Due
to small synthetic characters these errors are shown and
Figure 12 illustrates performance comparison of the pro- are similar to some handwritten characters. Thus the pro-
posed classifier with other classifiers. It can be seen that posed method has 42.15 and 0.3% faster than Zheng et al.
the proposed classifier has high classification rate of [30] and Lin et al. [31], respectively, and highly accurate.
94% for handwritten text identification. While compar- Due to the presence of a large number of features Zheng
ing with other techniques the classification time is less in et al. [30] consume more time than the proposed method.
the proposed than in K-NN.
5. CONCLUSION
The comparative analysis of the segmentation using
Modified Pihu method and existing method is shown in This paper proposes a new SCA algorithm for identifi-
Table 3. Additionally the average execution time of the cation of handwritten Marathi character using WOAR-
Modified Pihu method as compared with the existing SVM-based classification. The OCR system performance
method to segment the natural image was about 2.72s can be influenced by many factor varieties. The recogni-
which was 55.16% lesser than existing methods [26]. tion system uses the Modified Pihu method for charac-
Table 4 shows the performance of the proposed algorithm ter segmentation process. From the segmented Marathi
in terms of error rates and classification time. character various features are extracted and the accu-
racy of our proposed method is enhanced by utiliz-
Marathi character. The proposed approach accomplishes ing these features. The SCA algorithm is used to select
high accuracy, precision, recall, F-measure, FAR and FRR the best solutions, after the SCA process the opti-
and further the error rates are less while comparing with mized feature vectors are given to the input of the
existing techniques for accurately recognizing the sub- WOAR-SVM classifier to classify and recognize the best
stantial characters. handwritten.
12 S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION

REFERENCES Information Systems for Indian Languages, pp. 224–227,


Berlin, Heidelberg: Springer, 2011.
1. P. Kamble, and R. Hegadi, “Handwritten Marathi character
recognition using R-HOG feature,” Procedia. Comput. Sci.,
14. S. Farkya, G. Surampudi and A. Kothari, “Hindi speech
Vol. 45, pp. 266–274, Jan.2015.
synthesis by concatenation of recognized hand written dev-
nagri script using support vector machines classifier”, 2015
2. R. Hegadi, and P. Kamble. “Recognition of Marathi Hand-
International Conference on Communications and Signal
written Numerals Using Multi-Layer Feed-Forward Neu-
Processing (ICCSP), Apr.2015.
ral Network”, 2014 World Congress on Computing and
Communication Technologies, Feb.2014.
15. M. Hanmandlu, and O. Murthy, “Fuzzy model based recog-
nition of handwritten numerals,” Pattern Recognit., Vol. 40,
3. N. Garg, L. Kaur, and M. Jindal, “Segmentation of touching no. 6, pp. 1840–1854, Jun.2007.
modifiers and consonants in middle region of handwritten
Hindi text,” Pattern Recognit. Image Anal., Vol. 25, no. 3, 16. C. Yang, and Y. Yang, “Improved local binary pattern for
pp. 413–417, Jul.2015. real scene optical character recognition,” Pattern Recognit.
Lett., Vol. 100, pp. 14–21, Dec.2017.
4. G. Pagare and K. Verma, “Associative Memory Model for
Distorted On-Line Devanagari Character Recognition”, 17. R. Sarkhel, N. Das, A. Das, M. Kundu, and M. Nasipuri, “A
2015 Fifth International Conference on Advances in Com- multi-scale deep quad tree based feature extraction method
puting and Communications (ICACC), Sep.2015. for the recognition of isolated handwritten characters of
popular indic scripts,” Pattern Recognit., Vol. 71, pp. 78–93,
5. P. Roy, A. Bhunia, A. Das, P. Dey, and U. Pal, “HMM-based Nov.2017.
Indic handwritten word recognition using zone segmenta-
tion,” Pattern Recognit., Vol. 60, pp. 1057–1075, Dec.2016. 18. Y. Zhang, S. Liang, S. Nie, W. Liu, and S. Peng, “Robust
offline handwritten character recognition through explor-
6. S. Connell, and A. Jain, “Template-based online charac- ing writer-independent features under the guidance of
ter recognition,” Pattern Recognit., Vol. 34, no. 1, pp. 1–14, printed data,” Pattern Recognit. Lett., Vol. 106, pp. 20–26,
Jan.2001. Apr.2018.

7. H. Al-Muhtaseb, S. Mahmoud, and R. Qahwaji, “Recogni- 19. X. Xiao, L. Jin, Y. Yang, W. Yang, J. Sun, and T. Chang,
tion of off-line printed Arabic text using hidden Markov “Building fast and compact convolutional neural networks
models,” Signal. Processing., Vol. 88, no. 12, pp. 2902–2912, for offline handwritten Chinese character recognition,”
Dec.2008. Pattern Recognit., Vol. 72, pp. 72–81, Dec.2017.

8. S. Naz, K. Hayat, M. Imran Razzak, M. Waqas Anwar, S. 20. Y. Wu, F. Yin, and C. Liu, “Improving handwritten Chinese
Madani, and S. Khan, “The optical character recognition text recognition using neural network language models
of Urdu-like cursive scripts,” Pattern Recognit., Vol. 47, no. and convolutional neural network shape models,” Pattern
3, pp. 1229–1248, Mar.2014. Recognit., Vol. 65, pp. 251–264, May.2017.

9. A. Choudhary, R. Rishi, and S. Ahlawat, “Off-line hand- 21. R. Pramanik, and S. Bag, “Shape decomposition-based
written character recognition using features extracted handwritten compound character recognition for Bangla
from Binarization technique,” AASRI Procedia, Vol. 4, pp. OCR,” J. Vis. Commun. Image. Represent., Vol. 50, pp.
306–312, Jan.2013. 123–134, Jan.2018.

10. S. Pasha, and M. Padma. “Handwritten Kannada char- 22. R. Mouhcine, A. Mustapha, and M. Zouhir, “Recogni-
acter recognition using wavelet transform and structural tion of cursive Arabic handwritten text using embed-
features”, 2015 International Conference on Emerging ded training based on HMMs,” Journal of Electrical Sys-
Research in Electronics, Computer Science and Technol- tems and Information Technology., Vol. 5, pp. 245–251,
ogy (ICERECT), Dec.2015. Apr.2017.

11. J. Olszewska, “Active contour based optical character 23. C. Boufenar, A. Kerboua, and M. Batouche. “Investigation
recognition for automated scene understanding,” Neuro on deep learning for off-line handwritten.
Computing, Vol. 161, pp. 65–71, Aug.2015.
24. S. Roy, N. Das, M. Kundu, and M. Nasipuri, “Handwritten
12. P. Singh, R. Sarkar, N. Das, S. Basu, M. Kundu, and M. isolated Bangla compound character recognition: A new
Nasipuri, “Benchmark databases of handwritten Bangla- benchmark using a novel deep learning approach,” Pattern
Roman and Devanagari-Roman mixed-script document Recognit. Lett., Vol. 90, pp. 15–21, Apr. 2017.
images,” Multimed. Tools. Appl., Vol. 45, pp. 11–19,
May.2017. 25. J. Sueiras, V. Ruiz, A. Sanchez, and J. Velez, “Offline contin-
uous handwriting recognition using sequence to sequence
13. G. Verma, S. Prasad and P. Kumar, “Handwritten Hindi neural networks,” Neuro Computing, Vol. 289, pp. 119–128,
character recognition using curvelet transform”, May.2018.
S. P. RAMTEKE ET AL.: A NOVEL WEIGHTED SVM CLASSIFIER BASED ON SCA FOR HANDWRITTEN MARATHI CHARACTER RECOGNITION 13

26. S. Ramana Murthy, V. Roy, M. H. Narang, and S. 29. S. Mirjalili, “SCA: A sine cosine algorithm for solving
Gupta, “An approach to divide pre-detected Devanagari optimization problems,” Knowl. Based. Syst., Vol. 96, pp.
words from the scene images into characters,” Signal. 120–133, Mar.2016.
Image. Video. Process., Vol. 7, no. 6, pp. 1071–1082, Nov.
2012. 30. Y. Zheng, H. Li, and D. Doermann, “Machine printed text
and handwriting identification in noisy document images,”
27. K. Jindal, and R. Kumar, “A new method for segmentation IEEE Trans. Pattern Anal. Mach. Intell., Vol. 26, no. 3, pp.
of pre-detected Devanagari words from the scene images: 337–353, Mar.2004.
Pihu method,” Comput. Electr. Eng., Vol.70, pp.754–763,
Dec.2017. 31. Y. Lin, Y. Song, Y. Li, F. Wang, and K. He, “Multilingual
corpus construction based on printed and handwritten
28. S. Rajashekararadhya and P. Ranjan, “Zone based feature character separation,” Multimed. Tools. Appl., Vol. 76, no.
extraction algorithm for handwritten Numeral recogni- 3, pp. 4123–4139, Feb.2015.
tion of Kannada script”, 2009 IEEE International Advance
Computing Conference, Mar. 2009.

Authors Dhiraj S. Deshmukh received B.E degree


in Mechanical Engineering from Y.C.C.E.
Surendra P. Ramteke received B.E degree (PTDC), Nagpur affiliated to R.T.M.
in Electronics and Telecommunication Nagpur University Nagpur and M.Tech.
Engineering from B.N.College of Engi- Degree in M.Tech. (HPE) from V.R.C.E.,
neering Pusad Dist.Yavatmal in 2002 and Nagpur presently NIT, Nagpur. He pur-
M.E. Degree in Digital Electronics from sued Ph.D. in Mechanical Engineering
Prof. Ram Meghe Institute of Technology from R.T.M. Nagpur University, Nagpur
& Research, Badnera Amravati affiliated in 2012. He is presently working as Associate professor in
to Sant Gadge Baba Amravati University Mechanical Engineering at G H Raisoni College of Engineer-
Amravati, Maharashtra in 2011. He is currently working as an ing, Nagpur, Maharashtra, India. He has published 78 research
Assistant Professor and pursuing Ph.D. degree at the Depart- papers in International Journals,17 research papers in Interna-
ment of Electronics & Telecommunication Engineering, SSBTs tional Conference. His research interests includes Heat Power
College of Engineering &Technology Bambhori Jalgaon, Maha- Engineering.
rashtra, affiliated to North Maharashtra University Jalgaon
India. He is life Member of ISTE. His areas of interests include
Image Processing and Pattern recognition.
Corresponding author. Email: rsurendra0711@gmail.com
Ajay A. Gurjar belongs to the Depart-
ment of Electronics and Telecommu-
nication, Sipna College Of Engineering
And Technology, Amravati Maharashtra,
India. After having procured Bachelor’s
Degree in Electronics & Telecommuni-
cation in1995, he completed his Mas-
ter’s Degree in the same subject in 2003.
Presently he is working as a Professor and
Dean Academics in Sipna College of Engineering & Technol-
ogy, Amravati.

You might also like