| Résumé |
For the analysis and the transformation of speech signals, we develop a new method to estimate the glottal flow by inversion of the vocal tract. However, no measure of this flow exist in vivo presently. It is thus very difficult to validate a such estimation. Thanks to high-speed videoendoscopy, we can estimate the glottal area. It allows us to compute a theoretical glottal flow by the use of a physical model (eg. S.Maeda). The glottal flow estimated by the inversion of the vocal tract and the flow obtained by the glottal area can be finaly compared. Therefore, we made a method of automatic measure of the glottal area on the videoendoscopic images. The vocal folds are filmed through a rigid endoscope which passes through the mouth, connected to a high-speed camera ENDOCAM 5562 which provides 4000 color images a second in 256x256 pixels. The glottal area is estimated by a thresholding method of the luminance of the videoendoscopic images. The threshold is automatically computed by localising the edges of the vocal folds. We will describe the method of glottal area estimation and finish by a comparison of the various speech signals: acoustic, Electro-Glotto-Graphic, glottal area and glottal flows. |
