Speech segmentation and classification using higher order moments