Arabic character recognition using spiking neural network

Mohd Said, Muhammad Lutfi Mirzan (2023) Arabic character recognition using spiking neural network. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full text) Arabic character recognition using spiking neural network.pdf - Submitted Version Download (3MB)

Abstract

Spiking Neural Network is a new generation which is the third generation of architecture neural network. This network is based on a biological mechanism that encodes the information using a pulse or spike. It has advantages in terms of training speed compared to the previous generation. This project uses a handwriting Arabic character dataset that has 28 characters. There are challenges because the handwriting Arabic character does not have many datasets, which is difficult to train. The size is not the same, and it is also different from other characters because of Arabic writing in cursive. When written, it will have many different writing styles, even if it is the same letter. Therefore, this project will increase the dataset using the augmentation process and enhanced pre-process to resize so that all sizes are the same and spiking neural networks can apply. Next, this project will validate the accuracy and compare it with another neural network model that uses a handwritten Arabic dataset. So, on this project, BindsNET will be use as the main framework, and then it will use PyCharm software to run this project. Handwritten Arabic dataset from Kaggle will be used to train Arabic characters. In conclusion, at the last of this project spiking neural network can be applied for Arabic character recognition

Actions (login required)

View Item