We describe and analyze a discriminative algorithm for learning to align a phoneme sequence of a speech utterance with its acoustical signal counterpart by predicting a timing sequence representing the phoneme start times. In contrast to common HMM-based a ...
John Wiley and Sons2009