A hierarchical Dirichlet process for the background interference suppression to improve the microphone array imaging results

Aerodynamic and aeroacoustic measurements in the wind tunnel are essential for the structural design and optimization of the aircraft. Microphone arrays are widely used in wind tunnels for the identification of aircraft noise sources, where the arrays are exposed to strong background interference. Meanwhile, the background interference is varied in different non-anechoic chamber acoustic experiments, and background interference directly affects the identification of noise sources. A generalized array denoising algorithm is proposed to address the interference problem. A hierarchical Dirichlet process is developed to deal with background interference that has non-independent and non-identical distribution characteristics between different microphone channels. At the same time, the sound source signal is also modelled based on its low-rank characteristic. All involved parameters in the model are estimated by the variational Bayesian algorithm. Then, the source signal can be separated from complex background interference. The denoising algorithm is also applied in simulations and wind tunnel experiments to verify its effectiveness and robustness in suppressing complex background interference suppression.