Automatic segmentation of speech recorded in unknown noisy channel characteristics