The Arabic Optical Characters Recognition (AOCR) is one of the challenging recognition tasks nowadays, as Arabic handwriting is cursive and contains many dots. Dots are a big challenge for Arabic recognizers, as writers sometimes connect them. Moreover, Dots are prone to be considered noise. This paper proposes a new divide and conquers based approach that tries to conquer dots problems. The novelty of the proposed approach is in its feature extraction method. The extracted features are used to train Artificial Neural Network (ANN) Feed-forward. The result is interesting and shows that this method should be further investigated.