A Novel Method For Speech Segmentation Based On Speakers' Characteristics