Application of Speaking Style Conversion in the Czech and Slovak TTS System with Cepstral Description

This contribution describes experiments with speaking style conversion performed on the utterances produced by the TTS system with cepstral description and basic prosody generated by rules. Speaking style conversion was realized in the form of post-processing operation on the output speech signal of the TTS system, and as a real-time implementation directly to the TTS system. Speaking style prototypes representing three emotional states (sad, angry, and joyous) were obtained from the sentences with the same information content. The problem with the different frame length between the prototype and the target utterance was solved by linear time scale mapping. The results were evaluated by listening tests of the resynthetized utterances.