Handwritten Cursive English Text Recognition Using Deep CNN-RNN based CT

Abstract

The Cursive Handwriting Recognition is employed in applications like reading aid for the blind, recognition of historical documents, processing of bank cheques, pattern recognition then on which defines ability of a system to spot characters. These systems transcribes human written texts into digital texts. Back propagation algorithm, binary segmentation algorithm, Optical Character Recognition systems are commonly used for translating any handwriting into plain text format. the main drawback of those techniques is that their efficiency is a smaller amount. Hence neural networks are widely adopted for classification and performance approximation tasks. during this proposed technique, offline HWRs is completed using CNN-RNN and Tensorflow - Keras Functional API .The proposed technique used segmentation free approach to create a system ready to">which can be able to recognize the handwritten characters with highest accuracy. The proposed DCNN-CTC are able to do character error rate of 4.07%, yielding a relative CER reduction of 30.8%. therefore the main motive of the project is that we are getting to convert the unrecognizable cursive handwritten texts into an easier text.