Offline Handwritten Numeral Recognition Using Convolution Neural Network
Abhisek Sethy, Prashanta Kumar Patra, Soumya Ranjan Nayak · 2020
In this current digital age of world, character recognition (CR) has been done through various machine learning algorithms. And it considered to be one the most challenging segment of pattern recognition. In addition to the above context, offline handwritten character is the most challenging one as compared with the printed one. Despite various algorithms that were harnessed on various handwritten scripts, it can be possible to have more feasibility solution and high recognition rate. Here, in this paper, we have focused on the handwritten numerals of Odia and Bangla scripts. To overcome the ambiguities that arise in handwritten, one has been resolved using the Convolutional Neural Network (CNN). Here we have suggested a state-of-the-art CNN-based approach for recognition of multiple handwritten numerals of both the scripts and clearly shown how effectively it has been used for evaluating the discriminate features from the original image and later leads to report high recognition rate. At the simulation level, we have listed up variance nature of the individual's images, and through CNN, a high recognition rate is achieved, which is quite helpful in building the automatic recognition system for handwritten numerals to have solution for real-time problems.