Design and simulation of handwritten detection via generative adversarial networks and convolutional neural network

In the field of Computer Vision, optical character recognition (OCR) plays a crucial part. In OCR, handwritten document images are transformed into machine-encoded text. Handwritten character recognition is an essential research field of pattern recognition and has attracted broad studies. Convolutional Neural Network (CNN) is a popular choice used with deep learning, in recognizing handwritten characters and many research papers have already been published in this area. Implementation of neural network in different forms is found in most of the work and produces different accuracy levels due to complex letter structure, data unavailability, and different writing styles. We can increase the accuracy of recognising theEnglish handwritten character by our proposed method, producing new training instances from the existing instances using augmentation and Generative Adversarial Networks (GANs). We use a deep learning model in this article, GAN for generating English handwritten character recognition. Furthermore, we illustrate that how generated characters can be used to increase the performance of English handwritten characters classification. With the help of this proposed model, we attained the accuracy of 95.36% for recognising characters.