Paper Title
Deep Learning Approach for Offline Handwritten Character Recognition
Abstract
Character recognition is one of the main aspects of computer vision. It is a technology that provides full
alphanumeric representation of printed and handwritten characters. Many researchers have proposed various methods and
approaches for character recognition. Neural network method performs well for pattern classification and is utilized both for
feature extraction and classification. This paper aims to build an offline handwritten character recognition (HWCR) using
deep neural network (DNN). EMNIST Letters and Balanced dataset are used in training and testing phase for character
classification. The proposed deep neural network uses three hidden layers and a softmax layer. The performance of DNN is
derived from the total no of characters correctly detected. The performance of DNN is compared with logistic regression and
shallow network method.
Keywords - Deep Neural Network, EMNIST, Computer Vision, Handwritten Character Recognition (HWCR).