Othcr a framework for offline tamil handwritten character recognition using deep learning approaches

Abstract

In common, handwritten Tamil scripts would be recognized and newlinethere is need to convert into editable digital format from scanned handwritten newlinecharacter image. The scribbled characters of the Tamil language are difficult newlineto recognize due to variances in shape, position of the object, size, similar newlineshape introduced by writers, unnecessary curves, character discontinuity and newlineunwanted loops. Due to the error occurring state this recognizing system is newlinemore complex. The most of inaccuracies are caused by the anarchy with newlinesimilar shapes. To address these drawbacks, digital image of a manuscript newlinehave to be documented as digitally by editable format. Numerous researches newlinebased on offline Tamil recognition deals only with few Tamil characters since newlineit becomes extremely complicated in distinguishing small variations in large newlinehandwritten documents. The writer s complexity affects the overall formation newlineof the characters. Such types of complexities are due to discontinuation of newlinestructures, unnecessary over loops, variation in shapes as well as irregular newlinecurves. These complex issues result in enhanced error value rates. Therefore, newlineto overcome such issues, this article proposes a novel approach to enhance the newlineoffline Tamil handwritten character recognition by utilizing four principal newlinesteps: Pre-processing, Segmentation, Feature extraction and classification. For newlineoptimal segmentation of Tamil characters, this paper utilizes the Tsallis newlineentropy approach based atom search (TEAS) optimization algorithm. Then a newlineNewton Algorithm (NM) based deep convolution extreme learning (DELM) newlineapproach is utilized for the extraction and classification of input images. newlineFinally, the experiments are carried out for numerous Tamil handwritten newlinerecognition based approaches. The proposed Tamil character recognition newlineutilizes the datasets of Isolated Tamil handwritten characters established by newlineHP lab India to evaluate the efficiency of the system. The proposed DCELM newlineapproach to recognize then predict the isolated characters to overcome the newlinedifficulties in offl