ACCENTS Transactions on Image Processing and Computer Vision, Vol 6(21)
ISSN (Online): 2455-4707
http://dx.doi.org/10.19101/TIPCV.2020.618051
68
Deep ensemble neural networks for recognizing isolated Arabic handwritten
characters
Haifa Alyahya
1*
, Mohamed Maher Ben Ismail
2
and AbdulMalik Al-Salman
3
Lecturer, Department of Computer Science, King Saud University, Saudi Arabia
1
Associate Professor, Department of Computer Science, King Saud University, Saudi Arabia
2
Professor, Department of Computer Science, King Saud University, Saudi Arabia
3
Received: 20-August-2020; Revised: 25-October-2020; Accepted: 05-November-2020
©2020 Haifa Alyahya et al. This is an open access article distributed under the Creative Commons Attribution (CC BY) License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1.Introduction
The widespread use of the Arabic language and the
vast number of Arabic books and documents in
several Arab and international countries has
promoted researchers' interest in handwritten
characters recognition [1]. In particular, these efforts
were mainly intended to digitalize a large number of
historical Arabic documents and ease the exploitation
of the knowledge they enclose.
The Arabic Handwritten Character Recognition
(AHCR) problem has been formulated as a typical
pattern recognition problem. Specifically, as depicted
in Figure 1, the earliest AHCR systems include
components shared by various image-based pattern
recognition applications such as medical image
classification, cancer detection, biometric, etc. The
main purpose of the pre-processing phase is to
eliminate the effectiveness of noise, writing size,
style differences, and duplicated points [2].
*Author for correspondence
The five main steps of pre-processing are size
normalization, de-hooking or interpolating missing
points, smoothing, slant correction or duplicated
point removal, and resampling of points, as shown in
Figure 2.
Figure 1 Main steps of a typical character
recognition systems
Input
Pre-processing
Segmentation
Feature extraction
Recognition
Output
Research Article
Abstract
In recent years, handwritten character recognition has become an active research field. In particular, digitalization has
triggered the interest of researchers from various computing disciplines to address several handwriting related
challenges. Despite these efforts, there are still opportunities for the development and improvement of the recognition of
the handwritten Arabic letters. In this paper, we designed and developed a deep ensemble architecture in which ResNet-
18 architecture is exploited to model and classify character images. Specifically, we adapted ResNet-18 by adding a
dropout layer after all convolutional layer and integrated it in multiple ensemble models to automatically recognize
isolated handwritten Arabic characters. A standard Arabic Handwritten Character Dataset (AHCD) was used in the
experiments to train and assess all the proposed models. Satisfactory results were obtained using all models. The best-
attained accuracy was 98.30% using a typical ResNet-18 model. Similarly, 98.00% and 98.03% accuracies were obtained
using an ensemble model with one fully connected layer (1 FC) and an ensemble with two fully connected layers (2 FC)
coupled with a dropout layer, respectively.
Keywords
Handwriting character recognition, Arabic, OCR, Online recognition, Handwritten recognition.