A method comprising receiving an audio input signal comprising speech, determining an accent class corresponding to the speech, identifying an accented phone pattern within the speech, replacing the accented phone pattern with an unaccented phone pattern, and generating an unaccented output signal from the unaccented phone pattern.