Research On Spectral Enhancement Technology For Ship Radiated Noise

The core of underwater battlefield acoustic signal perception lies in the detection and recognition of ship radiated noise.Ship radiated noise consists of line spectra and continuous spectra,with line spectra exhibiting better stability and higher intensity.They are important research targets in passive detection and recognition,carrying significant research significance.In order to obtain line spectral information more effectively,a series of measures need to be employed.This paper focuses on the research of adaptive enhancement,detection,and extraction techniques related to line spectrum.Firstly,this paper investigates the fundamental theories of line spectral enhancement,detection,and extraction.Based on the basic theory of adaptive line spectral enhancement,the performance of conventional adaptive line spectral enhancers is simulated and analyzed under different input signal-to-noise ratios and multi-line spectral conditions.The results show that the conventional adaptive line spectral enhancer exhibits performance degradation and loss of line spectra when the signal-to-noise ratio falls below a certain threshold.To address the issue of average power spectrum gain loss caused by phase differences,coherent average power spectrum is studied.The results indicate that coherent average power spectrum reduces the variance of power spectrum estimation and improves the performance by approximately 3d B.Based on the peak characteristics of line spectra and three criteria for line spectrum extraction,both fixed threshold and automatic threshold extraction methods are investigated.Simulation results demonstrate that the automatic threshold method exhibits stronger line spectrum extraction capability,especially in conditions with significant noise background fluctuations.Next,based on the research on conventional adaptive line spectral enhancement in the previous chapter,further investigations on adaptive line spectral enhancement techniques were conducted in this paper.By analyzing the frequency-domain sparsity of steady-state adaptive weights,it was concluded that the frequency domain of optimal adaptive weights exhibits strong sparsity,laying the foundation for subsequent sparsity improvements.The sparse-driven adaptive line spectral enhancer was studied,and simulation results indicated that as the input signal-to-noise ratio increases,the frequency-domain sparsity of the sparse-driven adaptive line spectral enhancer gradually becomes stronger and approaches the optimal weights.Addressing the limitation of the ₁l-norm-based improved sparse-driven adaptive line spectral enhancer’s inability to achieve more ideal line spectral enhancement,a regularization technique was extended to the range of 0 to 1,resulting in a sparse-driven adaptive line spectral enhancer based on the l_1/2-norm.Simulation results demonstrated that the l_1/2-norm-based improved sparse-driven adaptive line spectral enhancer exhibits stronger weight sparsity and superior performance.Lastly,in order to fully utilize the spatial-temporal domain information and further enhance the algorithm’s performance,this paper investigates the adaptive line spectral enhancement technique based on spatial-temporal domain integration.Addressing the performance degradation of ALE under low input signal-to-noise ratio conditions,a novel approach is proposed in this study,which employs the spatial-temporal domain integration filter as a pre-processing module for the sparse-driven ALE.Simulation analysis and experimental results from at-sea trials demonstrate that the spatial-temporal domain integration-based sparse-driven adaptive line spectral enhancer effectively adapts to enhance line spectral signals in the spatial-temporal domain while reducing steady-state errors.This approach overcomes the issues faced by conventional adaptive line spectral enhancers,such as susceptibility to steady-state adaptive weight noise and performance degradation under low input signal-to-noise ratio conditions.Moreover,it imposes lower signal-to-noise ratio requirements and significantly improves the performance of line spectral enhancement.