Automatic Arabic handwritten text recognition is still an open research field, methods that describe satisfactory solution are still lacking. This can be attributed to cursive orthography and to the letter shape context sensitivity, which complex the problem of the character segmentation. This paper presents a heuristic rule based analytical segmentation approach for handwritten Arabic text, which preceded by a pre-process phase that handles binarization, short gaps closing, skew estimation, and critical features points calculation. Unlike other approaches a broader set of candidates for segmentation is generated and multi phase election process is performed to elect the best suitable candidates. Experiments are conducted on a database of 50 images of text sentences with an average of 4 words. Results were very satisfactory and outperform literature documented results.