This paper discusses pattern recognition techniques for classification of speech into voiced and unvoiced sounds. An algorithm is described which is capable of adapting to changing speaker characteristics, spectral distortion and background noise. Robust speech parameters are used in a computationally efficient multi-dimensional classifier which requires no training data base.
Bibliographic reference. Moulsley, T. J. / Holmes, P. R. (1989): "An adaptive voiced/unvoiced speech classifier", In EUROSPEECH-1989, 1466-1469.