Development of Handwriting Recognition System in Postal Service Sector

Abstract

Handwriting recognition is a comparatively popular research due to its diverse applicable environment. It helps to solve complicated problems and at the same time, it reduces manpower consumption. This paper proposes a system for recognizing online handwritten characters by using KNearest Neighbor (KNN). General steps of an algorithm are: (1) capturing the postcode and name of district area by using external web camera, (2) performing image processing on the image, (3) creating input data for KNN by extracting vital feature from each character, (4) classifying the dataset using KNN algorithm and performing recognition during the test, and (5) providing result of the recognition. The experiment was carried out in the aspect of text font size, the density of text and light intensity of background text. Experiment results show that training sets, trained inputs and untrained inputs achieved reasonably good result with an accuracy rate of 100%, 87.54% and 75.35% respectively. For processing time, the training sets consumed the lowest processing time which is 195.32ms, followed by trained inputs with 201.30ms and untrained inputs with 204.98ms. Additionally, medium font size, high-density text and optimum intensity of the background text managed to achieve high accuracy rate and low processing time. In this way, the system is able to help the postal services sector to speed up the sorting process as well as reducing manpower consumption in the sorting unit at the same time. Overall, the system has fulfilled the objective of the project, which is to propose high accuracy and short processing time of the handwriting recognition system.