Letter Recognition from Noisy Images Using Deep Learning

Abstract

In many fields, both machine learning and deep learning are being used. One of the most active and difficult areas of research is letter recognition. Recently, deep learning and letter recognition have drawn the attention of many researchers. Letter recognition is a problem that has been worked on in many languages. Letter recognition refers to the process of identifying and distinguishing letters based on their visual features. It involves analyzing patterns of responses and detecting specific components that form different letters. This perceptual activity plays a crucial role in recognizing individual letters and is often studied using rapid visual displays and visual search tasks. Additionally, letter recognition in context explores how letters are processed within words and contributes to our understanding of reading processes. In the process of recognition, pre processing techniques increase image quality by reducing noise and correcting orientation, while convolutional neural networks extract letter features. However, the existing letter recognition system faces many challenges in extracting text from noisy and distorted images or complex layouts, and extraction is mostly limited to numbers and the English alphabet. So, several studies using deep learning models have been conducted to achieve better accuracy. The proposed technique achieves 96.2% accuracy in recognizing letters from input images.