ISCA Archive Eurospeech 1989
ISCA Archive Eurospeech 1989

A time-domain articulatory speech synthesiser

F. J. Owens, G. T. H. Wright, N. W. Ramsey

This paper describes a new time-domain articulatory speech synthesis model. The system consists of a model of the speech articulators, a time-domain simulation of a lossy acoustic tube model of the vocal tract and a two-mass model of the vocal cords. The vocal tract area function is expressed as a transformation of ten articulatory variables which provide a two-dimensional geometrical description of the vocal tract in the mid-sagittal plane. The acoustic tube model is based on the Kelly-Lochbaum structure which uses the method of reflection coefficients to compute the foward and backward sound pressure waves in a lossless digital transmission line analogue of the vocal tract. The basic Kelly-Lochbaum structure has been extended to permit the modelling of viscous friction losses, yielding wall losses, radiation from the mouth, variable vocal tract length, aspiration and frication.