Perception of CV-Syllables Isolated From Japanese Connected Speech

Abstract

From the standpoint of connected speech recognition, experiments were performed to investigate the perceptual nature of CV-syllables removed from Japanese connected speech. The recognition of individual vowels and syllables was found to be seriously impaired when they were taken out of their phonetic environments and presented as monosyllables to listeners. The recognition, however, substantially improved as the speech segments were lengthened. Finally, it was found that a vowel and a CV-syllable were completely recognized when they were in the middle position of a speech unit consisting of three successive CV-syllables. The results indicate that at least two syllables, one preceding and one following, are necessary to provide an acoustic environment for the correct identification of a CV-syllable.