Principal components analysis for Hindi digits recognition

Author

El-Bashir, M.S. ; Rahmat, Rahmita Wirza O K ; Ahmad, Fatima ; Sulaiman, Md Nasir

Author_Institution

Univ. Putra Malaysia, Serdang

fYear

2008

fDate

13-15 May 2008

Firstpage

738

Lastpage

740

Abstract

The recognition process depends on the how features are extracted. There are several ways for feature extraction but the most important is to extract the most effective features and can distinct between patterns. In this research, an approach is proposed to recognize Hindi numerals. Initially image is enhanced and normalized. After that, PCA is applied for feature extraction. Recognition is performed by using first and second Norm. Another two more norms were proposed named ENorm and EEuclidean. Results showed 93.5%, 94.79%, 95% and 94.79% recognition accuracy when applying first norm, ENorm, second norm and EEuclidean respectively.

Keywords

character recognition; feature extraction; image enhancement; natural languages; principal component analysis; Hindi digit recognition; feature extraction; image enhancement; image normalization; principal component analysis; Computer science; Data mining; Feature extraction; Handwriting recognition; Image databases; Information technology; Matrix converters; Pattern recognition; Principal component analysis; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer and Communication Engineering, 2008. ICCCE 2008. International Conference on

Conference_Location

Kuala Lumpur

Print_ISBN

978-1-4244-1691-2

Electronic_ISBN

978-1-4244-1692-9

Type

conf

DOI

10.1109/ICCCE.2008.4580702

Filename

4580702