A hybrid Feature Selection Approach for Arabic Handwritten Text Based on Genetic Algorithm and Support Vector Machine

Reducing features dimensionality in Arabic handwritten text recognition is a crucial issue, due to its impact on the overall recognition performance. In this paper, an Arabic handwritten text recognition model based on a hybrid feature-selection approach and Artificial Neural Network (ANN) classifier is proposed. The proposed feature selection approach mainly based on Genetic Algorithm (GA) and Support Vector Machine (SVM). By doing so, we aim to reduce the dimensionality of Arabic handwritten text features and optimize the recognition overall performances. The proposed based on GA-SVM approach was tested using the IFN/ENIT dataset. The results of the proposed features-selection are promising; since they reduce the number of tested features up to 78% by removing the irrelevant and redundant features. Using the ANN classifier, the proposed model reaches a 96.5% recognition rate, which supports its effectiveness in recognizing Arabic handwritten texts.