At our laboratories we are presently developing an automatic labelling method. In our approach, knowledge of the behaviour of signal parameters has been formalised in the form of context-sensitive rules and stored in a knowledge base. Prior to the labelling, the rules take a given phonetic transcription as an input and produce a suggestion of what the actual labelling procedure should be looking for in the speech signal. In the present contribution an outline of our knowledge based approach will be given.
Bibliographic reference. Houben, C. G. J. (1989): "Automatic labelling of speech using an acoustic-phonetic knowledge base", In EUROSPEECH-1989, 2104-2107.