Interspeech'2005 - Eurospeech
This paper presents ProZed, an environment for the multilingual analysis by synthesis of speech prosody. The analysis is based on the symbolic representation of prosodic form without reference to prosodic function. The parameters of the model are at present limited to fundamental frequency and duration but the same framework could be extended to accomodate other parameters such as spectral tilt or voice quality. Each parameter is defined with respect to specific domains and units for long-term and shortterm phonetic interpretation with an abstract annotation system corresponding to a level of surface phonological representation. The environment is integrated with the Praat program for analysis and the Mbrola program for synthesis.
Bibliographic reference. Hirst, Daniel / Auran, Cyril (2005): "Analysis by synthesis of speech prosody: the Prozed environment", In INTERSPEECH-2005, 3225-3228.