Character and Digit Recognition Using ANN Back Propagation Algorithm and Image Segmentation
Sadaf Waziry, Ahmad Bilal Wardak, Jawad Rasheed · 2022 International Conference on Smart Information Systems and Technologies (SIST) · 2022
In the field of pattern recognition, one of the demanding areas of study is character and digit recognition. It has broad use, including car license plate identification, bank account numbers, postal codes, invoice numbers, etc. In this work, an effort is made to recognize letters and numbers from images for English letters and numbers with an image segmentation process using a multilayer feedforward neural network. The data used for this work contains 36 classes which 10 belong to numbers and 26 for letters. 80% of the data is used for training the neural network model and 20% is used for testing the algorithm. In this work, the images are resized for training into 28x28 pixels, each of which has 784 pixels that will be used as features for neural network algorithm training. The final outcome shows that there are a strong recognition rate and 95% of accuracy for this work.