A new text independent method for phonema segmentation