Texture Feature Local Binary Pattern for Handwritten Character Recognition

Handwriting is still widely used as a research topic in computer vision topics. Two important parameters to consider are classification performance and classification time. Local Binary Pattern (LBP) is a method of extracting shape features in handwriting, but the number of features generated is quite large, resulting in increased classification time. This research proposes a method to speed up the classification time by dividing the LBP features into several grids. It is found that the proposed method can provide good performance at 90.99% accuracy and sufficient speed, on LBP parameter [33], 5x5 grid and KNN parameter k=3.

I. Introduction

Handwriting is still widely used as a research topic, including in computer vision topics such as digital image processing, image recognition, and image understanding. Especially in today’s digital world, where the use of paper documents is minimized. In the case of image recognition, machine learning must be made that can be supervised or unsupervised. In the supervised method, the actual class of the observed data is known [1]. Whereas in the unsupervised method, the actual class of the observed data is unknown.