Physics-informed neural network for predicting in vacuo vocal fold eigenmodes: A proof of concept study

This study investigates a machine-learning approach for real-time computation of in vacuo vocal fold eigenmodes. A physics-informed neural network is trained to predict eigenmodes and eigenfrequencies by integrating the governing equations of vocal fold dynamics. The proposed architecture predicts physically consistent modal shapes and frequency estimates, showing strong agreement with finite-element method results for lower-order modes, achieving mean relative errors below 6% in eigenfrequency prediction and cosine correlation values approaching 1 in eigenmode prediction, while prediction accuracy declines for higher-order modes. These findings demonstrate that the physics-informed neural network provides an accurate and efficient real-time computation of in vacuo vocal fold eigenmodes.