Danish Speech Synthesis (DST project)



Speech and Multimedia Communication The Danish Research Ministry is funding a 27-month project that will develop a new generation highly intelligible and natural-sounding text-to-speech synthesizers for Danish. The synthesizer is expected to gain high market penetration through high quality combined with attractive pricing. The consortium behind the project comprises: The Institute for General and Applied Linguistics, Copenhagen University; TeleDanmark A/S; Speech-Ware A/S, and the Center for PersonKommunikation, Aalborg University. CPK's main responsibility within the project is to carry out research on and develop techniques for synthetic speech signal generation. Furthermore, CPK co-ordinates activities on assessment. The current signal generation module is based on Residual Excited Linear Prediction technique for time scaling, pitch modifications and unit concatenations. At the moment two Danish databases are released: a male and a female voice. Both databases consist of approx. 3000 speech units (di- and triphones) extracted from carefully pronounced nonsense words. The final version of the text-to-speech synthesiser was delivered to the Danish Research Ministry in due time November 31 2001. The improvements in intelligibility and naturalness as compared to an earlier version were documented through a formal assessment session. For further information about research within this area and online demonstrations refer to: http://www.cpk.auc.dk/syntese/ and http://www.speech-ware.com. This year's publications comprise [Barry, Nielsen and Andersen, 2001; Hoequist and Andersen, 2001a-c; Nielsen, Hoequist and Andersen, 2001; Andersen, 2001a-c; Andersen, Danielsen and Hoequist, 2001]. (Ove Andersen, Charles Hoequist, Paul Dalsgaard)
Effektiv start/slut dato31/12/200131/12/2001