Recognition of Cursive Roman Handwriting - Past, Present and Future
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 1
Benefit of multiclassifier systems for Arabic handwritten words recognition
ICDAR '05 Proceedings of the Eighth International Conference on Document Analysis and Recognition
Offline Arabic Handwriting Recognition: A Survey
IEEE Transactions on Pattern Analysis and Machine Intelligence
Classifiers combination and syntax analysis for Arabic literal amount recognition
Engineering Applications of Artificial Intelligence
| Hi-index | 0.00 |
In this paper, we present a global recognition system for Arabic handwritten words; we focus on the two phases of feature extraction and classification. In our system, we have retained three feature sets. The Zernike moments and the structural features of the word are extracted from the binary image, the Freeman code is established from the contour image of the word and the zoning is given from the skeleton image. These features, representing the words, are extracted to be used as input, in an individual or combined way, of the four classifiers used in our system: the Fuzzy C-Means algorithm (FCM), the K-Means algorithm, the K Nearest Neighbor algorithm (KNN) and a Probabilistic Neural Network (PNN). The system architecture is a parallel one where each expert (classifier) gives his point of view and we combine the results to make a final decision. The classifier results are combined using two methods: the simple vote and the weighted sum.