Efficient Recognition of Bangla Handwritten Digits Based on Deep Neural Network
Md. Lizur Rahman1, Ifrat Jahan2, Akash Saha3, Md. Nawab Yousuf Ali4
1Md. Lizur Rahman, East West University, Dhaka, Bangladesh.
2Ifrat Jahan, East West University, Dhaka, Bangladesh.
3Akash Saha, East West University, Dhaka, Bangladesh.
4Md. Nawab Yousuf Ali, East West University, Dhaka, Bangladesh.
Manuscript received on 24 September 2018 | Revised Manuscript received on 30 September 2018 | Manuscript published on 30 November 2018 | PP: 200-203 | Volume-7 Issue-4, November 2018 | Retrieval Number: E1828017519©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Nw-a-days world has started to move into machine based technologies. Recognition of various features, shapes, images etc., has become extremely excited topics over recent years. Many authors proposed various techniques to recognition of handwritten digits on different languages. This paper presents a new technique based on deep neural network for the purpose of efficiently recognition of handwritten digits for Bangla language. Two datasets are used in this paper including CMATERDB 3.1.1 dataset and ISI (Indian Statistical Institute) dataset. About 24500 samples are used for training purpose and 4800 samples are used for testing purpose and the proposed technique achieves 98.70 percent accuracy. This paper also presents detailed overview on artificial neurons, and deep neural network. In addition, the efficiency of proposed method shown by comparing the results with other existing techniques.
Keywords: Handwritten Digit Recognition; Perceptron; Sigmoid Neuron; Dropout; Deep Neural Network; Artificial Neuron; Bangla Digit.
Scope of the Article: Image Processing and Pattern Recognition.