Improved Handwritten Digit Recognition Using Deep Learning Technique

Abstract

Handwritten digit recognition (HDR) is a fascinating field of research with practical applications in various domains. Imagine automatically processing checks, deciphering handwritten notes, or interacting with devices using intuitive scribbles - this is the potential of HDR.HDR tasks a computer with understanding the nuances of human handwriting, a seemingly simple yet surprisingly complex endeavor. Unlike standardized fonts, individual handwriting styles exhibit unique characteristics, making recognition a challenging feat.Variations in pressure, slant, size, and even individual loopsand strokes all contribute to the individuality of handwritten digits. Despite these challenges, HDR research continues to evolve, with deep learning techniques playing a crucial role
in recent advancements. This paper explores the state-of-theart in deep learning-based HDR and proposes an innovative approach to address the aforementioned challenges. In this Paper, to evaluate CNN’s performance, we used the MNIST dataset, which contains 70,000 images of handwritten digits. Achieves 98.2% accuracy for handwritten digit. And where 40 of the total images were used to test the data set