An experimental software system has been developed to perform syllable-based speech synthesis. The synthesis units are demisyllables with affixes, whereby the demisyllables consist of a vowel portion and a consonant cluster. For the synthesis of unrestricted German an inventory of about 1300 demisyllables is needed. The system allows the modification of demisyllable duration, amplitude and pitch. Parameters can be interpolated at the transition between consecutive demisyllables. Experiments have been carried out to establish demisyllable concatenation rules. A medial consonant cluster can be composed of a final and an initial consonant cluster. Using this method the majority of coarticulation effects of fluent speech can also be handled. The synthesis is carried out by an LPC-vocoder using PARCOR-coefficients.
[1]
John E. Markel,et al.
Linear Prediction of Speech
,
1976,
Communication and Cybernetics.
[2]
Günther Ruske,et al.
An approach to speech recognition using syllabic decision units
,
1978,
ICASSP.
[3]
Marian J. Macchi.
A phonetic dictionary for demisyllabic speech synthesis
,
1980,
ICASSP.
[4]
O. Fujimura,et al.
A demisyllable inventory for speech synthesis
,
1979
.
[5]
Catherine P. Browman.
Rules for demisyllable synthesis using Lingua, a language interpreter
,
1980,
ICASSP.
[6]
Victor Zue,et al.
Diphone synthesis for phonetic vocoding
,
1979,
ICASSP.