Classifiers combination and syntax analysis for Arabic literal amount recognition

Abstract

Automatic handwriting recognition has a variety of applications in real world problems, such as mail sorting and check processing. Recently, it has been demonstrated that combining the decisions of several classifiers and integrating multiple information sources can lead to better recognition results. This article presents an approach for recognizing handwritten Arabic literal (legal) amounts. The proposed system uses a set of holistic structural features to describe the words. These features are presented to three classifiers: multilayer neural network, k nearest neighbor, and fuzzy k nearest neighbor. The classification results are then combined using several schemes; we retained the score summation one for this work. A syntactic post-classification process is then carried out to find the best match among the candidate words. The performance of this approach is superior to the system which ignores all contextual information and simply relies on the recognition scores of the recognizers.