An enhanced quadratic angular feature extraction model for Arabic handwritten literal amount recognition

Abstract

Arabic script has a number of characteristics that makes it unique among other scripts. Several feature extraction methods use statistical pixel distribution-based approach to recognize handwritten digits and words. These methods produce features that provide low complexity and high speed in terms of extraction performance. Angular feature extraction method, a pixel distribution-based, estimates the angular span features from the whole image depending on the center of gravity. This method was successfully used with Arabic (Indian) numbers but not with Arabic handwritten words. In this paper, we propose an enhanced quadratic angular feature extraction model, as a new statistical feature extraction model to recognize Arabic handwritten word used in bank cheque. AHDB standard dataset was used to evaluate the proposed model and the experimental results were compared with the previous studies conducted on the same dataset. The results show that the recognition rate was 59% with 15% enhancement than the previous works that used pixel distribution-based methods. Moreover, the combination between the proposed model and the perceptual model (PFM) has achieved outstanding results with recognition rate of 83.06%