Comparative Study on Handwritten Digit Recognition Classifier Using CNN and Machine Learning Algorithms
Tanuja Kumari, Yatharth Vardan, Prashant Giridhar Shambharkar, Yash Gandhi
Abstract
Digit Recognition is essential for interpreting image processing and pattern recognition since a machine cannot classify handwritten digits. Many real-time applications include OCR (Optical Character Recognition), which recognizes characters and digitizes printed texts. Converting handwritten digits to digital characters has been a challenging problem since the past. The physical documents cannot be efficiently processed without converting them to digital copies and it requires a lot of time and efforts. To provide a solution to handwritten classification, several algorithms and techniques have been proposed over the years. The objective of this research is to use Convolutional Neural Networks (CNN), K-Nearest Neighbor, and Support Vector Machine to recognize isolated handwritten digits. After implementing and training the models on the same dataset and comparing the results obtained for three different models, the results show that CNN is the most optimal machine learning technique to classify handwritten digits with an accuracy of 99.59 percent.