Syllable affixes in speech snythesis

Fujimura has proposed that English syllables be treated as a syllable core plus syllable affixes [O. Fujimura, J. Acoust. Soc. Am. 59, S25 (A) (1976)]. Since the core may contain only one true postvocalic consonant, syllable-final consonant clusters are derived by attaching one or more affixes, the apical consonants /tdszθ/. According to this analysis, it should be possible, in a speech synthesis-by-rule system, to concatenate syllable affixes to the syllable core with only simple coarticulatory rules. We report on several experiments testing the feasibility of this approach. In one experiment, phonetic affix segments corresponding to /s/ and /z/ were sliced onto CVC syllable cores, in which the core-final consonant varied in terms of place and manner of articulation. The original and concatenated versions of each test syllable were presented to listeners who rated each version according to naturalness. In general, there was no difference in rated naturalness between the concatenated syllables and the originals, showing the feasibility of the process of affixation for /s/ and /z/. A similar experiment is being conducted for /t/ and /d/ affixes.