Abstract
Speech production measurements have revealed many regularities in the time domain .
The literature has focussed on systematic variations in vowel durations , as these
are most likely to carry perceptually relevant information . Yet, not much attention
has been paid to the actual relevancy of vowel duration variations to speech perception.
It is our hypothesis that the complex system of temporal regularities
present in speech production can be reduced to a more general, limited set of rules
to be used in speech synthesis, which will prove to be equally acceptable in speech
perception . An experiment has been set up to test the influence of temporal variations
in the vowel /a:/ in isolated, diphone-concatenated words, depending on the
postvocalic consonant, syllabic structure of the word and presence or absence of
accent. Results show that a limited set of rules can be used to generate acceptable
synthetic speech .
Original language | English |
---|---|
Pages (from-to) | 33-42 |
Journal | IPO Annual Progress Report |
Volume | 20 |
Publication status | Published - 1985 |