DocumentCode :
2074483
Title :
A Persian Writer Identification Method Based on Gradient Features and Neural Networks
Author :
Ram, Soheila Sadeghi ; Moghaddam, Mohsen Ebrahimi
Author_Institution :
Electron. & Comput. Eng. Dept., Shahid Beheshti Univ., Tehran, Iran
fYear :
2009
fDate :
17-19 Oct. 2009
Firstpage :
1
Lastpage :
4
Abstract :
Newly, the effectiveness of Gradient features has been verified for writer identification of Latin texts. However, no researches on the performance of these features on Persian handwritten have been reported. Special styles of Persian handwritten assert different approaches to identify the writer in compare with other alphabets. This paper introduces a text-independent Persian writer identification method that its simplicity and accuracy is due to use two items: Gradient features that are abstracted for Persian documents, and Neural Network as a classifier. The results showed that Gradient features gained a satisfactory identification rate. The accuracy of system was about 94% for 250 handwritten samples from 50 writers.
Keywords :
gradient methods; handwriting recognition; neural nets; gradient features; identification method; neural networks; writer identification; Computer networks; Feature extraction; Gabor filters; Genetic algorithms; Multilayer perceptrons; Neural networks; Spatial databases; Support vector machine classification; Support vector machines; Testing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Image and Signal Processing, 2009. CISP '09. 2nd International Congress on
Conference_Location :
Tianjin
Print_ISBN :
978-1-4244-4129-7
Electronic_ISBN :
978-1-4244-4131-0
Type :
conf
DOI :
10.1109/CISP.2009.5301092
Filename :
5301092
Link To Document :
بازگشت