Combining Neural Networks for Arabic Handwriting Recognition

Chergui Leila1, Kef Maamar2, and Chikhi Salim3
1Department of Computer Sciences, University Larbi Ben Mhidi, Algeria
2Department of Computer Sciences, University Hadj Lakhdar, Algeria
3Department of Computer Sciences, University Mentouri, Algeria

 

Abstract: Combining classifiers is an approach that has been shown to be useful on numerous occasions when striving for further improvement over the performance of individual classifiers. In this paper we present a Multiple Classifier System (MCS) for off-line Arabic handwriting recognition. The MCS combines three neuronal recognition systems based on Fuzzy ART network used for the first time in Arabic OCR, multi layer perceptron and radial basic functions. We use various feature sets based on Tchebichef, Hu and Zernike moments. For deriving the final decision, different combining schemes are applied. The best combination ensemble has a recognition rate of 90,10 %, which is significantly higher than the 84,31% achieved by the best individual classifier. To demonstrate the high performance of the classification system, the results are compared with three research using IFN/ENIT database.

  

Full Text