Prosodic enhancements for a musical object oriented formant synthesizer

This paper describes a real time object oriented formant synthesizer with musical capabilities. These capabilities originate from the support of the SKDNL and MIDI specifications. Apart from the music protocols, the synthesizer uses quantized variability for the fundamental frequency of the uttered phonemes in order to accurately depict intonation characteristics. The desired performance of the synthesizer is evaluated on the grounds of 'naturalness' and prosodic 'exactness'. The former has to do with the depth of analysis for the synthesized phonemes and the latter with the ability to describe intonation phenomena traced in the context of the supported languages. Thus far, the synthesizer can perform sufficiently in English (US) and Greek. In order to describe phenomena like petasi or vareia (accent of grave tone) that have been detected in Greek vocal forms, the synthesizer is equipped with a user interface that describes these phenomena, compiles and quantizes them in pitch-duration vectors and feeds in the synthesizer with the adequate information in order to produce as natural as possible utterances of synthetic speech.