International Journal of Computer Applications (0975 – 8887) Volume 31– No.3, October 2011 48 A Structured Analytical Approach to Handwritten Marathi vowels Recognition Nilima P. Patil K. P. Adhiya Surendra P. Ramteke SSBT’S College of Engineering & Technology Bambhori, Jalgaon ABSTRACT: In OCR domain, it is now widely accepted that a single feature extraction method and single classification algorithm can’t yields better performance rate. It is therefore, a compound feature extraction approach based on structural analysis for recognition of offline handwritten Marathi vowels is proposed. Though, Moment invariant technique is well experimented by many researchers, an attempt is made to enhance the existing results by extracting various supportive features like affine invariant moments, image thinning, structuring the image in box format, etc. These features are independent of size, slant, orientation, translation and other variations in handwritten characters. 5 samples of each vowel from 25 different people have been sampled and database was prepared. After segmentation, an individual image is resized to 50X50. 33 different features were evaluated for each character. The Fuzzy Gaussian Membership Function has been adopted for classification. The main objective of the paper is to test the possibility of using the MI, AMI combination of both for recognition of Handwritten Marathi vowels. The results show the satisfactory performance rate. Keywords: Feature Extraction, Moment invariants, OCR, Gaussian Function, 1. INTRODUCTION OCR has been extensively researched for more than four decades. With the advent of digital computers, many researchers and engineers have been involved in this important area. Handwritten Character recognition is an exigent task due to the restricted shape variation, different script style & different kind of noise that breaks the strokes in number or changes their topology [1, 2]. As handwriting varies when person write a same character twice, one can expect enormous dissimilarity among people. These are the reason that made researchers to find techniques that will improve the knack of computers to characterize and recognize handwritten characters [3, 4]. OCR is not only a new developing area due to many potential applications such as bank check processing, postal mail sorting, automatic reading of tax forms, and reading various handwritten and printed text and non-text documents[5,6]. It is also a benchmark for testing and verifying new pattern recognition theories and algorithms. It can be used as a reading machine for the visually handicapped when interfaced with a voice synthesizer [7, 8, 9]. Among the various phases of an OCR/HCR, the feature extraction phase is quite important as a set of useful properties of a character available as an image is defined and extracted during this phase. More relevant is the feature extraction method(s) used for discrimination [10, 11]. In this paper, we have studied, implemented and compared the performance of Moment Invariants based methodologies. 5 samples of each Marathi vowels from 25 different people have been sampled and database was prepared. After segmentation, an individual image is normalized to 50X50 pixel sizes. Seven moment invariants (MIs), Four Affine moment invariants (AMIs) & combination of both are evaluated for each character. Three kinds of moment invariants based features are extracted as follows i. Direct feature extraction. ii.Thinning process iii.Box approach. In each case we got seven features so total 21 features are evaluated. Similarly, these three kinds of features are evaluated by applying affine Moment invariant methods. So total 12 features and for combination of both methods, in all 33 features are used for processing of classification. The Fuzzy Gaussian Membership Function has been adopted for classification. The main objective of the paper is to compare the all feature extraction methods for recognition of Handwritten Marathi character independent of its Size, slant and other variations. The paper is organized as follows: In section 2, the approach of Moment Invariant, Affine moment Invariants of MI &AMI which is adopted for feature extraction, is presented. Section 3 deals with the classification based on Fuzzy Gaussian Membership Function. The section 4 gives details of result and conclusion is discussed in section 5. 1.1 Devanagari Marathi Vowels Devanagari, an alphabetic script, is used by a number of Indian Languages. It was developed to write Sanskrit but was later adapted to write many other languages such as Marathi, Hindi, Konkani, Sindhi and Nepali. Many other Indian languages use close variant of this script. Marathi script has about 11 vowels and 34 consonants. In addition, there are large number of conjuncts formed by combination of consonants, their half- forms and some modifier symbols. The modifier symbols are placed either on the top, at the bottom, on the left, to the right or a combination of these. Also has a Shirorekha (a header line) runs through the entire span of word. 2. FEATURE EXTRACTION 2.1 Moment Invariants Approach (MI)