Line Spectra Enhancement Technique Based on Auto-Adaptive Window Length

Feature extraction and parameter estimation of line spectra based on power spectrum is a practical issue frequently occurring in passive sonar application. It is considered the fact that the non-full period truncation of analytical data leads to frequency spectrum leakage which degrades feature extraction performance and parameter estimation precision of line spectra. To solve this problem or weaken its influence, the impact of non-full period truncation on signal-to-noise ratio (SNR) of line spectra is analyzed in this paper, and an iterative data-driven line spectra enhancement method based on auto-adaptive window length is proposed. The presented method has the capability of achieving line spectra enhancement and parameter estimation precision improvement, compared to the feature extraction method with fixed window length. The effectiveness of the proposed method is verified in the simulations and sea trial data.