(in lingua inglese)
In this paper, a recognition-based segmentation is proposed for online handwritten Arabic text. The strategy utilizes the dominant point detection algorithm with GA and HS to search for the best combination of segmentation points which scores the best recognition result. The strategy incorporates and modifies the dominant point detection algorithm to extract all possible segmentation points by transferring the text into a set of lines. Then, it locates the best segmentation points using a feedback system between GA and HS to find the minimal word score.