Physics-Informed Neural Operator for Speech Production Analysis
For speech researchers, this provides a fast, data-free simulator for vocal tract acoustics, though it is an incremental application of existing PINO methods to a new domain.
This work introduces the first physics-informed neural operator for speech production analysis, learning 1D wave equations without supervised data. It achieves 0.8% error in glottal volume flow and 3.2% error in speech waveforms compared to conventional methods, enabling GPU-parallelized simulation.
Physics-informed neural operators (PINOs) have recently gained attention as fast numerical simulators with potential for solving inverse problems. This study proposes the first PINO-based method for speech production analysis. The model learns the governing one-dimensional wave equations directly without requiring pre-computed supervised training data. Using vocal tract shape data as input features, we compare the proposed model's predicted f0, glottal volume velocity and sound pressure at the lip for five static vowels to a conventional Runge Kutta/Finite difference approach. With errors of 0.8% for glottal volume flow and 3.2% for speech waveforms, the proposed model enables efficient GPU-parallelized simulation without iterative calculations. We conclude that PINO is a promising approach for fast analysis of speech.