Sparse spatial-frequency adaptive learning for enhancing underwater acoustical tonals under impulse noise

Enhancing and extracting sinusoidal signals (tonals) is a key challenge in underwater acoustic signal processing. However, conventional tonal enhancement methods deteriorate seriously under high levels of impulse noise. To address this problem, we propose a sparse spatial-frequency adaptive learning method for robust tonal enhancement in the presence of impulse noise. An array signal model is formulated to characterize spatially correlated impulse noise. Based on this model, we exploit the sparsity of tonals in the spatial-frequency domain and develop a sparse spatial-frequency beamforming approach to achieve initial tonal enhancement. To mitigate the false alarms caused by impulse noise artifacts in the resulting azimuth-frequency spectrum, we further introduce an adaptive learning scheme. This scheme exploits the stationarity of tonal signals and the non-stationarity of impulse noise to effectively suppress noise-induced artifacts, thereby achieving a higher output signal-to-noise ratio. The proposed method is validated using both simulated and real experimental data. Comparative results demonstrate that our approach outperforms baseline methods in terms of azimuth resolution and processing gain under impulse noise conditions.