Two Robust Super Resolution Approaches in Aeroacoustic Imaging for Near-field Wideband Extended Sources

Recently deconvolution-based methods, like the DAMAS, have greatly improved spatial resolutions of the Beamforming in aeroacoustic imaging. But most of existing methods are not robust to background noise. In this paper, we propose two robust superresolution approaches using Sparsity Constraint (SC-RDAMAS) and Sparse Regularisation (SR-RDAMAS) respectively to simultaneously estimate source powers and positions, and the variance of background noise. In proposed SC-RDAMAS, sparsity constraint on source power is obtained by considering eigenvalue distributions of observed covariance matrix. When sparsity constraint is hard to determine in strong noise interference, proposed SR-RDAMAS applying l1 regularisation with proper regularisation parameter can greatly improve resolutions and robustness of proposed SC-RDAMAS. Moreover, proposed SC-RDAMAS can work well even if the source number is over-estimated, but our SR-RDAMAS does not require source number at all. Proposed methods are shown to be robust to noise, wide dynamic range, super resolution and feasibility to use for near-field wideband extended source imaging based on 2D non-uniform microphone array by simulated and wind tunnel data. Our methods are compared with the state-of-art methods: Beamforming, DAMAS, Diagonal Removal DAMAS, DAMAS with sparsity constraint, Covariance Matrix Fitting and CLEAN.