A Model For Handwriting Detection Using Graphology And Convolutional Neural Network (CNN)

Abstract

Detecting crime through handwriting can help reduce crimes involving handwriting, such as forgery, financial misappropriation, exam malpractice (where a paid individual writes an exam for a student), and document fortification. The problem associated with this research focuses on examination malpractice, where it is difficult to detect and identify one's handwriting from another, the aim of this research is to develop an Improved Handwriting Crime Detection Model using Graphology and Convolutional Neural Network (CNN). Python programming language was used to train the images with a Convolutional Neural Network (CNN), a total of 44 images were used in this research for training, 36 images for validation and 15 images were used to test the model. The model achieved an accuracy of 1.000 within a timeframe of 3secs and an overall performance of 98.7%. This model was able to detect and identify the handwriting of an individual from that of another individual.