Efficient 3D Acoustic Simulation of the Vocal Tract by Combining the Multimodal Method and Finite Elements

Acoustic simulation of sound propagation inside the vocal tract is a key element speech research, especially for articulatory synthesis, which allows one to relate physics production other fields science, such as perception. Usual methods, transmission line method, have very low computational cost and perform relatively good up 4-5 kHz, but are not satisfying above. Fully numerical 3D methods finite elements achieve best accuracy, high cost. Better performances achieved with state art semi-analytical they cannot describe geometry accurately fully (e.g. no possibility take into account curvature). This work proposes new method that achieves better description three-dimensional vocal-tract while keeping substantially lower than methods. It multimodal relies on two-dimensional compute transverse modes takes curvature variations cross-sectional area. The comparison simulations shows same degree accuracy (about 1 % difference in resonance frequencies) about 10 times lower.