Hand-Writing Recognition Using Structural, Statistical Features

Abstract

One of the very crucial challenges in pattern recognition operations is handwriting recognition, often known as handwritten number recognition. The processing of bank checks, the sorting of postal mail, the entry of data into forms, etc. are all procedures involving number recognition. The ability to create an efficient algorithm that can retrieve handwritten integers submitted by drug users via a scanner, tablet, and other digital gadgets is at the core of the issue. a method for reading handwritten numbers that are based on various machine-learning techniques. The major goal of this research is to provide reliable and efficient methods for handwritten integer recognition. Numerous machines learning techniques, including the Videlicet, Multilayer Perceptron, Support Vector Machine, Naive Bayes, Bayes Net, Scikit-Learn, Random Forest, J48, and Random. Using WEKA, trees have been used to recognize integers. This design mostly uses Scikit-Learn to celebrate handwritten integers. The scikit- learn library itself provides a wide variety of datasets for model training. In this design, an integer dataset has been utilized. A portion of the data will be fed into an SVC prophetic model, and the remaining data will be used to confirm the model's performance. It is also seen how the model's delicateness changes when the rate of training and test data changes.