Adaptive Line Enhancer for Passive Sonars Based on Frequency-Domain Sparsity, Shannon Entropy Criterion and Mixed-Weighted Error

作者：Zhe Li, Yusheng Cheng, Jiaxing Qiu · 发表于：Arabian Journal for Science and Engineering · 年份：2024 · DOI：10.1007/s13369-024-09682-3 · 被引用次数：8 · 研究领域：Advanced Adaptive Filtering Techniques、Speech and Audio Processing、Image and Signal Denoising Methods

Abstract Adaptive line enhancer (ALE) is one of the vital signal processing techniques to the detection and recognition of underwater acoustic targets for passive sonars. Conventional ALEs, based on Gaussian noise assumption and least mean square (LMS) algorithm, can achieve good line enhancement property in Gaussian noise background. However, limited by the high steady-state misadjustment of LMS algorithm, the performance of conventional ALEs deteriorates under non-Gaussian noise background and degrades severely in processing signals with comparably lower signal-to-noise ratio (SNR). Therefore, it’s of great necessity to improve the line enhancement performances of ALE techniques to meet the demands of engineering application in passive sonars. In order to optimize the robustness and adaptability of conventional ALEs in dealing with underwater acoustic signals with much lower-SNR and in non-Gaussian noise background, a modified ALE algorithm called frequency-domain ALE based on l 1 -norm, Shannon entropy criterion and mixed-weighted norm ( l 1 -SE-MWE-FALE) is proposed in this paper. The proposed l 1 -SE-MWE-FALE algorithm is based on the integration of frequency-domain sparsity, Shannon entropy (SE) criterion along with mixed-weighted error of LMS and least absolute deviation (LAD) to improve the ALE performance in situations above. The simulation results demonstrate that, when the input SNR is as low as – 25 dB, the local SNR (LSNR) gain for line spectrums by l 1 -SE-MWE-F...