المستخلص: |
In this paper, we describe a new approach for Off-line Arabic handwriting recognition using optimized multiple classifier system (MCS). Our strategy is based on Dynamic classifiers selection paradigm. It rests on proposed DECS-LR algorithm (Dynamic Ensemble of Classifiers Selection by Local Reliability) which is enriched the selection criterion by incorporating a new Local-Reliability measure and chooses the most confident ensemble of classifiers to label each test sample dynamically. On two databases sets using three fusion methods (voting, weighted voting), we can clearly show the effectiveness of the selection methods based on local accuracy estimation. We analyze that presented experimental results are encouraging and open other perspectives in the domain of classifiers selection especially speaking for Arabic Handwritten word recognition.
|