This paper deals with an acoustic analysis of the sets of Czech sentences uttered by single speaker. The data used in this analysis consists of both emotional and neutral sentences. We have been especially interested in some features which are supposed to influence the perception of speech, such as F0, phoneme duration, formant frequencies or energy. The analyzed sets of sentences were composed of utterances expressing various speaker's attitude. We have tried to reveal some acoustically measurable differences among various speaker's attitudes so that we could incorporate this knowledge into our speech synthesis system to obtain emotional synthetic speech.
展开▼
