Novel line spectrum frequency estimation method based on Mathematical Morphology filters

Line spectrum frequency estimation is very important to feature extraction for underwater target radiated noise. A novel method which could effectively reduce the influence of environment noises and find the position of line spectrum frequency on the power spectrum is studied in this paper. Dilation filter and Erosion filter, which are basic tools in Mathematical Morphology, are used to estimate the spectrum fluctuation range. The spectrum fluctuation range and the spectrum shape are two parts of which composing the estimation of the continuous spectrum. The line spectrum is obtained by subtracting the continuous spectrum from the power spectrum. Monte Carlo computer simulations are carried out to test the performance of the method proposed in this paper. The results verify that the estimation is accurate when the Line-spectrum-power-to-Continuous-spectrum-power Ratio (LCR) is above -5dB.