Recognition of Handwritten Bengali Characters using Low Cost Convolutional Neural Network
Rohit Jadhav, Siddhesh Gadge, Kedar Kharde, Siddhesh Bhere, Indu Dokare · 2022
Handwritten character recognition of Bangla Script is one of the most difficult and complex tasks in pattern recognition, because of the complicated alignment and similarity in the characters. This paper aims to explore the usage of a Convolutional Neural Network to recognize Handwritten Bangla characters. The classification stages and feature extraction stages of any pattern recognition task are responsible for accurately recognizing the patterns. The paper proposes a novel low-cost CNN architecture for Bengali Character Recognition over present well-known datasets CMATERdb, BanglaLekha-Isolated, Ekush. Using Convolutional Neural Network, the proposed model achieves good accuracy and well generalizes over multiple datasets. Overall Recognition accuracy obtained for datasets such as CMATERdb, BanglaLekha-Isolated, Ekush is 87%, 89.6%, 83.1% respectively.