Unsupervised Segmentation of Continuous Speech Using Vectorautoregressive Modeling