Acoustic modeling for robust utterance verification in continuous speech ...