Cross-speaker acoustic-to-articulatory inversion using phone-based trajectory HMM
Résumé
The article presents a statistical mapping approach for crossspeaker acoustic-to-articulatory inversion, i.e. estimating the most likely articulatory trajectories for a reference speaker from the speech audio signal of another speaker. This approach is developed in the framework of our system of visual articulatory feedback, developed for computer-assisted pronunciation training applications (CAPT). This system aims to provide any speaker with visual information about his/her own articulation, via a 3D talking head displaying all speech articulators. In the proposed approach, acoustic-to-articulatory inversion is achieved using a continuous feature mapping technique based on phonebased trajectory HMM. Speaker adaptation is addressed using a voice conversion approach, based on trajectory GMM.