Abstract:
Phoneme boundaries may be determined from a signal corresponding to recorded audio by extracting auditory attention features from the signal and extracting phoneme posteriors from the signal. The auditory attention features and phoneme posteriors may then be combined to detect boundaries in the signal.