This paper details the present state of a research programme aimed at setting up a multi-media acoustic-articulatory data base. The research is being carried out jointly as part of an 'acoustic-phonetic decoding' operation within the 'Speech Communication' section of the GRECO by two laboratories: the CERFIA, where homogeneous infra-phonemic segments are isolated by means of a pre-segmentation of the speech signal. The Institute of Phonetics at Aix-en-Provence, where the movements of the velum, the lateral walls of the pharynx and the lips during phonation are measured on pictures from video films. We will be describing the various processes used to establish a relation between the two sets of data and presenting the initial results of this confrontation.