Control of prosodic parameters for a formant synthesizer based on diphone concatenation

At the Deutsche Bundespost Research Institute, an audio response system (SAMT) based on the formant synthesis principle has been developed which is able to generate speech from sequences of monophone and prosody control codes. This paper describes patterns for controlling the fundamental frequency f0by means of which it is possible to reproduce the prosodic features stress, rising cadences and falling cadences. This is mainly achieved by three different f0patterns whose duration, amplitude Δ f0and position in the sentence depend on the length of the vowel in the potentially stressed (ictic) syllable and the sound environment of the vowel.