International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013
DOI:10.5121/ijfcst...
International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013
62
Preprocessing S...
International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013
63
3. PROPOSED CLA...
International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013
64
For instance, n...
International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013
65
Set Loop as tru...
International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013
66
Table 1. Recogn...
Upcoming SlideShare
Loading in...5
×

DEVNAGARI NUMERALS CLASSIFICATION AND RECOGNITION USING AN INTEGRATED APPROACH

257

Published on

Character recognition has always been a challenging field for the researchers. There has been an astounding progress in the development of the systems for character recognition. OCR performs the recognition of the text in the scanned document image and converts it into editable form. The OCR process can have several stages like preprocessing, segmentation, recognition and post processing. The recognition generally, consists of feature extraction and classification. The choice of features and classification scheme affects the performance of OCR largely. In this paper, a classification scheme is proposed for the Devnagari numerals, which forms the basis for recognition. This approach integrates the structural features and water reservoir analogy based feature to classify the Devnagari numeral. In order to classify a single numeral, at most four checks are required. This increases the efficiency of the proposed scheme.

Published in: Technology, Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
257
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
12
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

DEVNAGARI NUMERALS CLASSIFICATION AND RECOGNITION USING AN INTEGRATED APPROACH

  1. 1. International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013 DOI:10.5121/ijfcst.2013.3407 61 DEVNAGARI NUMERALS CLASSIFICATION AND RECOGNITION USING AN INTEGRATED APPROACH Kanika Bansal1 , Rajiv Kumar2 1 Student, SMCA, Thapar University, Patiala, Punjab, India. knbs_ind@yahoo.com 2 Assistant Professor, SMCA, Thapar University, Patiala, Punjab, India. rajiv.patiala@gmail.com ABSTRACT Character recognition has always been a challenging field for the researchers. There has been an astounding progress in the development of the systems for character recognition. OCR performs the recognition of the text in the scanned document image and converts it into editable form. The OCR process can have several stages like preprocessing, segmentation, recognition and post processing. The recognition generally, consists of feature extraction and classification. The choice of features and classification scheme affects the performance of OCR largely. In this paper, a classification scheme is proposed for the Devnagari numerals, which forms the basis for recognition. This approach integrates the structural features and water reservoir analogy based feature to classify the Devnagari numeral. In order to classify a single numeral, at most four checks are required. This increases the efficiency of the proposed scheme. KEYWORDS OCR, Devnagari Numerals, Feature Extraction, Classification. 1. INTRODUCTION Character recognition is used as an umbrella term, which covers all types of machine recognition of characters in various application domains. It is the processing of text based input patterns to produce meaningful outputs with the help of a machine. The input may come from online devices like tablets, stylus based devices or offline devices like scanners. Output may be a sequence of symbols like ‘Y’, ‘E’ ,‘S’ or a date on cheque like ‘Nov 14, 2011’ or validation result of a signature. This process can be used to translate articles, books and documents into electronic format, to publish text on website, to process the cheques in banks, to sort letters and many more applications. The image captured needs to pass through various stages which can be named as preprocessing, segmentation, recognition and post processing as given by [1,2]. The various stages of the character recognition process are shown in Figure 1 and are discussed here. Figure 1. Stages of Character Recognition Process [3]
  2. 2. International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013 62 Preprocessing Stage: The preprocessing stage includes all the functions required to produce the cleaned up version of image acquired through scanner. It includes the processes like filtering, binarization, thinning, smoothing. Segmentation Stage: The segmentation stage refers to decomposing the document into subcomponents. It deals with the separation of lines, words and characters according to the application through usage of various segmentation strategies as in [4]. Recognition Stage: The recognition stage, which uses the various pattern recognition strategies, assigns an unknown sample to a predefined class to identify the character according to the script used. Postprocessing Stage: The recognized characters are fed into the post processing stage, which includes the usage of dictionary according to the script for purpose of spell check, grammar check in order to enhance the rate of recognition. Since this paper talks about the feature extraction and classification, it is discussed in the following section. 2. FEATURE EXTRACTION AND CLASSIFICATION The recognition consists of two stages namely, feature extraction and classification. The recognition process assigns a character image to a predefined class by using a classification algorithm based on the features extracted and the relationships among the features, as in [5]. Feature extraction is concerned with the identification of the defining attributes of the character image. The selection of the features is a difficult task. As stated in [1], the features can be classified into statistical features and structural features. Statistical Features: It is a representation of a character image by statistical distribution of points. They mainly include features like zoning, projection, profiles, crossings and distances. Structural Features: They are based on topological and geometrical properties of the character. They mainly include features like aspect ratio, cross points, loops, branch points, strokes and their directions, horizontal curves at top or bottom etc. The extraction of good features is the main key to recognize an unknown character accurately. The classification stage is the main decision making stage of an OCR system and uses the features extracted in the previous stage to identify the text segment according to preset rules. The classification stage identifies each input character image by considering the detected features. The classification stage uses several techniques for achieving its objectives. The authors found that most approaches used for the classification are based on the neural networks and its adaptations. These techniques are computationally difficult and require good amount of time to perform the training of the systems in order to provide good results. Moreover, the accuracy of the recognition system depends on the quality of the training of the system. Thus, an effort has been made by the authors to develop a simpler approach for the character recognition process, which is discussed in the next section.
  3. 3. International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013 63 3. PROPOSED CLASSIFICATION AND RECOGNITION SCHEME The proposed scheme derives its basis from the concept and scheme described in [6]. The authors have developed the scheme for the Gurmukhi characters, where the water reservoir principle and features like number of components, presence of side bar or loop has been used to validate the segment of the input character image. They have used a hybrid scheme for the segmentation of the characters, which is used for the recognition in the proposed work. Since, the classification scheme works based on the features extracted from the character image, the features used in the proposed scheme are discussed first. The features used here are an integration of the statistical and structural features described in section 2. These features are explained in the next section. 3.1. FEATURES The features used by the scheme are presence of sidebar, presence of loop, presence of water reservoir. The loop and sidebar fall under the structural category of features while the water reservoir falls under the statistical category as it uses projection and profiles for its extraction. Thus, the current scheme is an integrated approach towards classification, which leads to recognition. The features used are discussed here:- Presence of Loop The presence of loop is very robust feature for classifying the characters. A closed loop is a pattern made up of several strokes that are formed when the writing instrument returns to a previous location while touching the writing surface continuously, giving a closed outline with a “hole” in the center. For instance, the numerals and have a loop while and does not have it. Presence of Sidebar The presence or absence of sidebar in the character is an important classification feature. The longest vertical run of black pixels on the columns of a text image can be said to a sidebar. For instance, the numeral has a sidebar, while and does not have it. Presence of Water Reservoir The presence of water reservoir and its type acts as a good classification feature. The water reservoir principle is as follows. If water is poured from one side of a component, the cavity regions of the component where water will be stored are considered as reservoirs. The top, bottom, left, and right reservoir of numerals are illustrated in Figure 2. Figure 2. Water Reservoirs in Numerals
  4. 4. International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013 64 For instance, numerals and have left reservoir, and have a top reservoir, and have a right reservoir. Based on these features, the classification structure is developed. The structure initiates with the checking of presence of reservoir in the character image. A character image can have more than one reservoirs present in it. The combination of the features extracted in the previous step, is used to classify an input image. The classification structure is as shown in Figure 3. Figure 3. Classification Structure for Devnagari Numerals Initially, the check for the presence of left reservoir is used to classify the characters into two sub classes. Then, the other features like presence of loop, presence of sidebar, presence of top and right reservoirs are used to classify them further. The check for presence of loop in the upper quadrant of the character image is also used to distinguish between two similar characters 2 and 3. The classification structure can be used to recognize the Devnagari numerals in an efficient manner. The proposed algorithm for the recognition of the Devnagari numerals based on the feature extraction and classification structure is stated in the following section. In the proposed work, the Top, Left, Right, Loop, Upper and Sidebar are Boolean variables. The Numeral is an integer variable to store the result. The algorithm runs on a character image. The proposed algorithm is:- Set Left as true if left reservoir is present in the character image Set Right as true if right reservoir is present in the character image Set Top as true if top reservoir is present in the character image Set Sidebar as true if sidebar is present in the character image
  5. 5. International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013 65 Set Loop as true if loop is present in the character image and Set Upper as true if loop is present in the upper quadrant of the character image If Left is true If Sidebar is true If Top is true Set Numeral as 5 Else Set Numeral as 1 End If Else If Top is true Set Numeral as 4 Else If Loop is true and Upper is true Set Numeral as 2 Else Set Numeral as 3 End If End If End If Else If Right is true If Sidebar is true Set Numeral as 8 Else If Loop is true Set Numeral as 6 Else Set Numeral as 9 End If End If Else If Top is true Set Numeral as 7 Else Set Numeral as 0 End If End If End If The proposed algorithm is implemented by the authors and the results obtained are discussed in the following section. 4. RESULTS The proposed algorithm for the classification and recognition of the Devnagari numerals is implemented by the authors and it is applied to several documents. The documents are referred as Doc1, Doc2, Doc3, Doc4, and Doc5 as shown in Table 1. The results are satisfying and summarized here:-
  6. 6. International Journal in Foundations of Computer Science & Technology (IJFCST), Vol. 3, No.4, July 2013 66 Table 1. Recognition Results for Handwritten Devnagari Numerals Document Numerals Present Numerals Recognized Accuracy (%) Doc1 10 10 100 Doc2 60 58 96.67 Doc3 20 20 100 Doc4 30 29 96.67 Doc5 25 24 96 5. CONCLUSION The Devnagari numeral recognition has application in various fields like reading details like postal zip code, employee code, passport number and processing of forms and bank cheques. These application demands an efficient scheme for the numeral recognition. The recognition scheme is composed of feature extraction and classification. The features used in the scheme are simple and easy to compute. The characters are initially classified on the presence of left reservoir into 2 sub classes. Further, the features presence of sidebar, presence of loop, presence of top and right reservoirs classifies the characters. The classification structure developed has a height of 4. Hence, in order to classify a numeral, at most 4 checks are needed. This reduces the efforts required to recognize the character, thereby, making it a simpler and easy task. Since, there is some inaccurate recognition, which can be attributed to the reason that some input images were not well formed. Hence, recognition results can be improvised if appropriate preprocessing is applied to the document image. REFERENCES [1] N. Arica and F. Yarman-Vural, (2001) “An Overview of Character Recognition focused on Off-line Handwriting,” IEEE Transactions on Systems, Man, and Cybernetics, vol. 31, no. 2, pp. 216-233. [2] A. Senior and A. Robinson, (1998) “An Off-line Cursive Handwriting Recognition System,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 20, no. 3, pp. 309-312. [3] R. Kumar and A. Dhiman, (2010) “Challenges in Segmentation of Text in Handwritten Gurmukhi Script,” in Proceedings of the International Conference on Recent Trends in Business Administration and Information Processing, CCIS 70, Springer-Verlag Berlin Heidelberg, pp. 388-392. [4] R. Casey and E. Lecolinet (1996) “A Survey of Methods and Strategies in Character Segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 18, no. 7, pp. 690-706. [5] G. S. Lehal and C. Singh (1999) “Feature Extraction and Classification for OCR of Gurmukhi Script,” Vivek, vol. 12, no. 2, pp. 2-12. [6] A. Kaur, R. Kumar and A. Singh (2010) “A Hybrid Approach to Classify Gurmukhi Script Characters,” International Journal of Recent Trends in Engineering and Technology, vol. 3, no. 2, pp. 103-105. AUTHORS Rajiv Kumar obtained his Ph.D. from the University College of Engineering, Punjabi University, Patiala, Punjab, India. At present, he is the Assistant Professor in the School of Mathematics and Computer Applications, Thapar University, Patiala. He has published several articles in international journals and conferences. Kanika Bansal obtained her B.Tech. (CSE) from Guru Gobind Singh Indraprastha University, Delhi, India and currently, pursuing her M.Tech. from Thapar University, Patiala, Punjab, India. She is currently working in the document image processing area.
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×