Face, Speech, and Acoustics |
||
Introduction · Program · Local Information · Registration · Contact |
For the most part of speech sounds their productions occur invisibly in the vocal tract other than labial articulations, that result not only in acoustic enhancement by lowering all formants but also in visible speech information.
In this talk I will present a simple two-dimensional lip model with 4 control parameters. The model needs as an input a range of formant values to output the regression coefficients for the calculation of the node coordinates of the graphic lip figure. The results of two linear regression analysis and that of the manual optimisation were shown. Two pure visual perception tests were carried out: one presentation of mouth photographs with teeth and tongue, and one presentation of lips only. The assumption was, that the information about backness/frontness were given by the teeth and tongue informations. The interesting results were discussed in the talk.