YorkTalk(c) Homepage



This home page gives some information on the YorkTalk speech generation system under development at the University of York UK.




WARNING

all the audio files (including the one for the spectrogram above)associated with icons on this page are *.aiff files (with .aif extensions) at 11025Hz sample rate. If your browser won't handle such files click on this icon to move to a collection of the same files with other audio formats.

END WARNING


Development history.

The audio files are in AIFF format with 11025Hz sample rate.

YorkTalk audio demonstrations and comparisons.

Sound icon click icon to listen to some isolated disyllables produced by the `industry standard' (segmental,formant synthesis) DECtalk system. This is included for comparison with the following YorkTalk example which synthesizes the same words. (196Kb)

Sound icon click icon to listen to some isolated disyllables produced by the YorkTalk nonsegmental speech generation system which employs the same Klatt synthesis engine as used to generate the DECtalk output of the previous example, but drives it with a different kind of knowledge about the linguistic structures involved and their phonetic interpretation. (183Kb)

The next sample provides an example of our recent, first (relatively successful) attempts to generate continuous speech.

click icon to hear an example of our (recent) 'rule'-generated continuous speech utterance. There are some artifacts in the output (noticeably in the control of frication and nasality). These arise because we have just `moved synthesisers'. Originally we employed the Klatt79 software synthesiser but, in order to gain more linguistically relevant control we have recently begun using Klatt's Klsyn88 synthesiser. In this latter synthesiser there are a number of individual gain controls and amplitude parameters which require slightly different balances from those we used previously. (99Kb)

Sound icon if you haven't already clicked on the picture, click icon to listen to the audio file which generates the YorkTalk spectrogram at the head of this page. The audio file was generated by the Klsyn88 formant synthesizer using a handcrafted parameter description file. The parameter file itself was created with `Iggy', the Interactive Graphics Editor.


Readme Icon Press to link to Information about our Interactive KLATT88 Parameter Editor developed by Dr Rob Fletcher of the Computing Service.



Click to go>
</A>  Back to John Local's Home Page
<p>
<hr>
<p>
Link to a form to leave <a href= comments or requests for information


YorkTalk is currently being developed by John Local, Richard Ogden and Steve Harlow.

Page constructed by John Local September 1994