Data-driven impedance tube method for prediction of normal sound absorption coefficient

The impedance tube method, particularly the two-microphone method, is widely employed for measuring normal acoustic parameters. However, this method is constrained by the assumption of plane wave, leading to limitations on the effective frequency range. To resolve the limitations, this paper proposes a data-driven impedance tube method capable of accurately predicting the normal sound absorption coefficient using only two microphones in a multi-modal field. First, the proposed method integrates a neural network model with the transfer relationship in the impedance tube, and generates a large number of pre-training datasets by constraining the boundary conditions of the physical model. Subsequently, the neural network is trained using a supervised learning strategy on these datasets to accurately learn the mapping relationship between the sound pressure vector and the amplitude vector. Finally, the predictive ability of the proposed method for the normal sound absorption coefficient has been verified by simulations and validated by experiments.