Generalized sound field interpolation for freely spaced microphone arrays in rotation-robust beamforming

In this paper, we present a novel method for aimed at achieving rotation-robust beamforming with circular microphone arrays (CMAs), where the microphone distribution is unknown. Previous methods require a known microphone distribution, limiting their applicability in real-world scenarios where such information is often unavailable. Our proposed method addresses this limitation by utilizing an unequally spaced circular microphone array (unes-CMA) with unknown microphone positions. The method comprises two key components: unsupervised calibration and unequally spaced . Unsupervised calibration employs an innovative iterative optimization technique to estimate the positional errors of the microphones without any pre-existing location information, thereby determining the microphone distribution on the unes-CMA in an unsupervised manner. Once these positional errors are estimated, the unequally spaced method enables the reconstruction of the target signal of the unes-CMA before rotation. This process of unsupervised calibration enables accurate even with unknown microphone distributions. Additionally, we further improve the method by modifying the cost function used during unsupervised calibration and extending its applicability to nearly circular microphone arrays, which are not strictly a circle anymore. Simulation experiments were conducted to evaluate the performance of our method. The results demonstrated that our method effectively mitigates the negative impact of unknown microphone placements. It yielded significant improvements in estimating the signal before rotation and the performance of beamforming compared with previous methods. • Novel method for rotation-robust beamforming on circular microphone arrays (CMAs). • Applicable to freely spaced CMAs with unknown microphone distributions. • Enhance applicability to microphone arrays which are not strictly a circle. • Proposed method outperforms previous methods in estimating signals before rotation. • Significant improvements in source enhancement compared with previous methods.