Offline Handwritten Character and Numeral Recognition: A Kernel-Based Approach

Introduction

Automatic Character Recognition has considered as being one of the emerging fields of Pattern Recognition. Henceforth, it was quite impressive for researchers to do more qualitative work to solve the real-world problems. Usually OCR is the procedure of conversion of physical character image to its digital form and which is very much helpful in document analysis. Apart from that it is software based phenomenon where it has wide scope of application in this current digital world, such as Banking sector, Post-office, Document processing etc. OCR system can be performed for both handwritten as well printed characters Mantas (1986). In each recognition system, it must include certain intermediate stages such as acquisition, preprocessing, feature extraction and last the classifications, all these must be well-defined in order to report high recognition rate Govindan & Shivaprasad (1990). According to researchers it has been reported that recognition of handwritten one is quite challenging as compared with printed one. In handwritten mode usually having some issue such that, different writing skill adopted by different person with different shape orientation as well reported by Pal & Chaudhuri (2004). Here in this article authors have established a recognition system for Odia Handwritten characters and numerals. Odia is one of the oldest scripts and the official language of Odisha state, India and usually spoken in eastern, western region of the country. To make a robust OCR system for Odia scripts is more challenging one that may be of as follows: i. All the characters obtained from individual writers are not same which may differs in size and shape. ii. Various writing style of characters make recognition challenging, iii. Maximum characters are round in shape, similar in size and overlaps. Hence, it adds more complex analysis to the recognition model for handwritten ones Saeys (2007). In general Odia scripts consist of 49 numbers of characters (consonant, vowels) and 10 numbers of numerals are present. Apart from it some conjunct characters are there which is combination one or more consonants. One of the advantages of the such type of scripts that it has no lower case and upper case of letters. The issue that arises during recognition of handwritten characters makes researchers to work more to provide robust techniques towards this problem domain. To have successive recognition accuracy for any OCR system one should put more emphasis on feature extraction technique.