Use of Reflection Coefficients of Burg’s Method for Improvement of Visual Support Way in Pronunciation Practice

Abstract

In pronunciation practice, the trainees essentially need interactive supports to progress. We have developed a pronunciation training system, where the lip posture and tongue shape that are estimated by the formants extracted by AR model using Burg’s method are shown to the trainee. Realization of real time operations is one of the most important problems in this system. However, formants estimation requires long computation time. In order to avoid spectrum estimation, we attempt to reconstruct these visual instructions from the reflection coefficients of Burg’s method. Here, we investigate relationship between the acoustic feature or articulation manner of the sound and the reflection coefficients. The vocal tract area function calculated from the reflection coefficient was confirmed to acquire feature of tongue shape extracted from MR image taken during uttering more precisely than the formant. This gave possibility to improve visual support way based on the formants.