Cursive Arabic handwritten word recognition system using majority voting and k-NN for feature descriptor selection
Soufiane Hamida,Bouchaib Cherradi,Oussama El Gannour,Abdelhadi Raihani,Hassan Ouajji
DOI: https://doi.org/10.1007/s11042-023-15167-6
IF: 2.577
2023-03-30
Multimedia Tools and Applications
Abstract:Handwriting text recognition for computer systems has been the subject of more and more research. Recognition of Arabic handwritten text is always an ongoing challenge, mainly due to the similarity between its letters and various writing styles. However, the problem of cursive handwriting recognition remains laborious due to the complexity of the Arabic handwriting morphology. This paper proposes a new efficient-based image processing approach that combines three image descriptors, the Oriented Gradient Histogram, the Gabor filter, and the Local Binary Pattern method for the features extraction phase. To prepare the training and testing datasets, we applied a series of preprocessing techniques on 100 classes selected from the handwritten Arabic database IFN/ENIT. Then, we trained the k-nearest neighbor algorithm (k-NN) to generate the best model for each feature extraction descriptor. The best k-NN model is used to classify Arabic handwritten images according to their classes. The model’s performance evaluation uses the main common metrics, namely accuracy, sensitivity, specificity, and precision. Based on the performance evaluation results of the three k-NN generated models, the Majority-voting algorithm is used to combine the prediction results. A high recognition rate of up to 99.88% is achieved, far exceeding the already state-of-the-art results on the IFN/ENIT dataset. The obtained results highlight the reliability of the final model generated for recognizing handwritten Arabic words.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering