Combination Classifier-Based Off-line Arabic Handwritten Word Recognition System

Offline text recognition is very important in a variety of applications, such as the automatic sorting of mail, altering ancient documents, and historical document analysis. This paper aims to develop an ensemble of various classifiers to enhance the recognition of Arabic handwritten words. To represent the handwritten words, features are generated by a histogram of oriented gradients and local binary patterns. where each of its feature dimensions is reduced using principal component analysis. The support vector machine classifier is given for each feature set separately. Two independent classifiers are to be produced. The outputs of the two classifiers are integrated using the Bayesian method. The AHDB database was used to test the suggested strategy. When the classifiers' outputs are combined, recognition rates improve and, in some circumstances, surpass those of cutting-edge recognition systems. The suggested strategy has produced noteworthy accomplishments compared with other studies.