A real-time system for visualizing tongue position during vowel phonation

Abstract

A real-time system to estimate and visualize the position of the tongue during vowel phonation is presented. The system uses Linear Predictive Coding (LPC) to track formants F1–F3 of input speech frames. An algorithm proposed by Ladefoged et al. is used to map formants F1–F3 to tongue position [J. Acoust. Soc. Am. 64, 1027 (1978)]. 2-D visualizations of the tongue are created with cubic splines, which are rendered each input frame to create a real-time animation. The proposed system could serve as a pedagogical tool for language learners and singers by providing visual feedback of the inner mechanics of the vocal tract during vowel phonation.