Modeling segment intonation for Slovene TTS system

A scheme for modeling the F/sub 0/ contour for different types of intonation units for the Slovene language is presented, for the purpose of constructing a Slovene text-to-speech (TTS) system. It is based on the results of analyzing F/sub 0/ contours using a quantitative model. Data from ten speakers was collected, resulting in a large corpus, mainly of declarative sentences. A method of generating the F/sub 0/ contour for a given utterance was defined, using only the text of the utterance as input. A near-to-natural synthesized F/sub 0/ contour was obtained by rules which regard the F/sub 0/ contour as the sum of global and local components.