Phoneme recognition based on top-down approach