Voice synthesis

The voice synthesis is the sound technique of Synthèse which aims to the artificial creation of words, via a system of Treatment of the signal. It can fall under a system of vocal interaction.

History

There were several techniques of voice synthesis until our days.

The first, popular between 1965 and 1985, is called voice synthesis by rules . This one is based on the modeling of the word starting from a sound Specter. Rules can be written to generate an artificial sound spectrum. This technique allows considerable savings in memory. The second technique, known as by concatenation of diphones , is not purely artificial. The synthesized sounds are in fact of the artificially attached segments of recording of word the ones following the others. This technique can be produced with less than 10 Méga sound bytes of data. The synthesized word seems more natural than that produced by rules, but of the problems persist when the length of Phonème, the intonation and the tonic Accent are taken into account. To cure these problems, it is possible to increase the quantity of sound extract to use for the Concaténation. Several extracts for same a diphone can be used in the same context, and several types of contexts (intonation, accent, type of sentence) can have each one their diphones. One speaks then about synthesis by selection of diphones or selection of units (Links Selection). Those are then used during the voice synthesis in order to reduce the bad transitions. This improvement can ask for databases of diphones of several Méga bytes, even of several Giga bytes.

Techniques

Formants

to see Formant

Intonation

to write

Diphones

A diphone represents the transition between two successive phonemes.

See too

External bonds

Demonstrations on line

  • Demonstration of the voice synthesis of SVOX

  • Demonstration of the voice synthesis by selection of unit of the European company Acapela Group

  • Demonstration of the voice synthesis of the american company Cepstral

  • Demonstration of the voice synthesis of Nuance (RealSpeak)

  • Demonstration of the voice synthesis of the Italian company Loquendo

  • Demonstration in line of Voice synthesis of SitePal

  • Demonstration of the voice synthesis of Multitel ASBL

  • Demonstration of the voice synthesis of Pediaphon (voice synthesis of the articles of French Wikipédia)

Software

  • Cepstral Swift (Windows and Mac OS X)

  • Infovox Desktop (Windows) and Infovox iVox (Mac OS X)

  • SnapVoice (Windows)

  • Digit PC (Windows)

  • Proloquo (Mac OS X)

  • GhostReader (Mac OS X)

  • Speechissimo (Mac OS X)

  • FreeTTS (Java)

  • Festival (Linux)

  • MBROLA (voice synthesizer)

  • elite (Windows & Linux)

  • eSpeak (Linux) free Voice synthesis for English and other languages.

  • DECtalk software commercial Voice synthesis (Linux) multilangues.

  • Sayz Me (Windows, free, simple of use, interfaces in English, possibility of adding voices for French)

  • free

    LIA_PHON {Voice synthesis (LPG) starting from unspecified text to interface with MBROLA.}

  • yread (free, Windows, compatible with protocol SAPI 5.1)

References

  • Voice synthesis and Voice recognition: Left right-hand sides and Parallel Worlds , T. Dutoit, L. Roofer, F. Malfrère, V Pagel, C. Laugh (http://tcts.fpms.ac.be/publications/papers/2002/cfa2002_tdlcfmvpcr.pdf)

Random links:Joystick | David De Freitas | Battle of Diu | Pontiac (provincial district) | Harold Cobert | Degli_Italiani_de_chant_de_l'IL