International Journal of Computer Applications (0975 – 8887) International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013) 12 Recognition of Gujarati Numerals using Hybrid Approach and Neural Networks Baheti M. J. Department of Computer Engineering, SNJB’s College of Engineering Chandwad, Nashik (M.S.) India Kale K. V. Department of CS and IT, Dr. BAM University, Aurangabad (M.S.) India ABSTRACT The handwriting recognition is the scheme of converting text symbolized in the spatial form of graphical symbols into its figurative depiction. Handwritten characters have been the most accredited technique of collecting, storing and transmitting information all the way through the centuries. To give the proper ability to the machine it requires studying the image-form of data which forms a special pattern to be interpreted. Designing and building machines that can recognize patterns remains one of the thrust areas in the field of computer sciences. A lot of work has been done in this field, but still the problem is not answered in its full density. A good text recognizer has many commercial and practical applications, e.g. from finding data in digitized book to computerization of any organization, like post office, which involve manual task of interpreting text. In this paper, we have presented a hybrid approach for recognition of Gujarati handwritten numerals using neural networks as classifier and achieved a good recognition rate for noisy numerals. Keywords Gujarati numerals, neural networks, handwriting recognition. 1. INTRODUCTION The handwriting recognition is the scheme of converting text symbolized in the spatial form of graphical symbols into its figurative depiction [1]. Till today, it remains one of the most challenging and exciting areas of research in computer science. In advanced years it has full-fledged into a grown-up stream of science, bringing into being a massive organization of work. Whilst the computer has hugely simplified the process of producing printed documents, the convenience of a pen and paper still makes it the natural medium for many important tasks. Since centuries, handwritten script has been the most approved method of collecting, storing and transmitting information. This is useful for making digital copies of handwritten documents, and also in many automated processing tasks, such as automatic mail sorting or cheque processing. In automated mail sorting, letters are directed to the correct location by recognition of the handwritten address. Similarly, cheque processing involves recognizing the words making up the cheque amount. This requires making the machines work like humans [2]. For giving machines the human like abilities, Artificial Intelligence, plays a very vital role. Giving machine the power to see, interpret and the ability to read text is one of the major tasks of Artificial Intelligence. Across India more or less twenty-four official languages are used from corner to corner within a variety of regions of the nation [3-7]. Each language has the diversity representing its eccentricity as well as carves up some of the resemblance with other languages. Apart from Gujarat, Gujarati is spoken and used as official language across globe within various countries like South Africa. Being derived from Devanagari, Gujarati language is written by using Gujarati script. Fig. (1) shows the numerals that belongs to Gujarati script. Fig. 1Gujarati numerals From Fig 1 it can be observed that Gujarati numerals show several identically similar numerals in Devanagari. This paper deals with the recognition of Gujarati handwritten numeral by hybrid approach using neural networks as classifier. This paper is organized in following sections; Section 2 describes brief literature survey done for Gujarati script recognition. Section 3 describes algorithm which we have used to implement the paper. Section 4 elaborates the neural network used for classification. Section 5 details the experimental results and successively section 6 concludes the work done. 2. LITERATURE SURVEY 2.1 Skew Detection and Correction Patel & Desai [8] had assumed that persons writing style remains uniform in most cases hence to speed up the image processing, they extracted middle part of the text to use as input for the radon transform for skew detection purpose. The radon transform computed projections of an image matrix along specified directions. The extracted middle portion of the text was considered as an input image for radon function. To detect the skew the radon transform for input image was computed at angles from 0° to 179°, in 1° increments of angles. The maximum value of R provided by the function [R,xp] = radon(I, theta); where I was input image and theta varied for 0 to 179, helped to detect the skew angle. The radon transform based techniques of skew detection and correction procured results on the skew angle in the range of -20 to +20 degrees. 2.2 Character Compression Yajnik [9] had proposed an approach of wavelet descriptors (Daubechies D4 wavelet coefficients) for image compression of printed Gujarati letters. The authors scanned the document at 300dpi. Daubechies D4 wavelet transform was applied to