Handwritten Tamil Character Recognition Using Convolution Neural Network by Adam Optimizer

Abstract

Optical Character Recognition is a widely used electronic method for recognition of handwritten images. Tamil handwritten character recognition is complex to recognize. Hence considerable research efforts have been taken in this field. The complexities of writers and the characters, structure over looping and unwanted character portions are the major challenges faced in Tamil characters. RGB to grayscale conversion, image complementation and structure morphing are enclosed in the preprocessing phase. The processed images are subject to recognition with optimized CNN. The connected layers are enhanced using ADAM optimizer for improvement of the standard. The accuracy and performance of the proposed work is compared with other models with certain performance measures.