Speech synthesis with accented diphones

A complete inventoly of accented and unaccented diphones, produced by a professional male speaker, was made. In order to evaluate the quality of synthetic speech using both types of diphones, we asked subjects to judge the naturalness and fluency of different versions of an utterance. Two experiments with isolated polysyllabic words indicate that the use of accented or unaccented diphones has a perceptual effect on long vowels only. In a third experiment, we evaluated the use of the different diphone types in short sentences with a fixed temporal struc ture. It appears that the use of unaccented instead of accented diphones for unaccented syllables does not result systematically in more natural-sounding speech.