Arabic Handwriting Recognition Using Concavity Features and Classifier Fusion

This paper presents a simple and effective technique for the recognition of writer-independent offline handwritten Arabic Digits. The system is based on labeling the white pixels in a digit's image into nine different concavity categories. Four different feature vectors are extracted from these labeled concavities. Each feature vector is then introduced to a linear SVM classifiers. The final decision of the system is achieved using classifiers fusion methods. The system has been tested on a database of 10000 Arabic handwritten digits. The presented method achieves a recognition rate of 99.36% which outperforms all reported results on that Arabic digits database using linear SVM classifier.