DocumentCode :
3231210
Title :
A very high accuracy handwritten character recognition system for Farsi/Arabic digits using Convolutional Neural Networks
Author :
Ahranjany, Sajjad S. ; Razzazi, Farbod ; Ghassemian, Mohammad H.
Author_Institution :
Dept. of Electr. Eng., Islamic Azad Univ., Tehran, Iran
fYear :
2010
fDate :
23-26 Sept. 2010
Firstpage :
1585
Lastpage :
1592
Abstract :
In this paper, a new method is presented for recognizing the handwritten Farsi/Arabic digits by fusing the recognition results of a number of Convolutional Neural Networks with gradient descent training algorithm. Convolutional Neural Networks are a type of neural networks that are biologically inspired from human visual system which combines feature extraction and classification stages. This paper is concentrated on two main contributions. The first one is automatic extraction of input pattern´s features by using a CNN for Farsi digits and the second one is fusing the results of boosted classifiers to compensate the recognizers´ errors. The difference between competing systems is in the training set, which the frequency of samples that are “hard to recognize” were become higher in boosted classifiers. In addition, two rejection strategies were proposed and evaluated to find out “hard to recognize” samples. The experiments were conducted on extended IFH-CDB test database. The results reveal a very high accuracy classifier outperforming most of the previous systems. The achieved result shows 99.17% in recognition rate. In addition, the result was grown up to 99.98% after rejection of ten percents of “hard to recognize” samples.
Keywords :
feature extraction; gradient methods; handwritten character recognition; image classification; natural language processing; neural nets; Arabic digit; Farsi digit; boosted classifier; convolutional neural network; feature extraction; gradient descent training algorithm; handwritten character recognition system; human visual system; Filtering; Humans; Pixel; Training; Convolutional neural networks; LeNet-5; automatic document management; gradient-based learning; optical character recognition;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Bio-Inspired Computing: Theories and Applications (BIC-TA), 2010 IEEE Fifth International Conference on
Conference_Location :
Changsha
Print_ISBN :
978-1-4244-6437-1
Type :
conf
DOI :
10.1109/BICTA.2010.5645265
Filename :
5645265
Link To Document :
بازگشت