Two ways of propeller recognition feature extraction

The analysis of the radiated-noise of vessels given in this paper shows some strong superposed line components in low frequency spectrum below 100OHz occurring at discrete frequencies which correspond with the rotation speed of propeller shaft, propeller blade frequency and their harmonic frequency. The line components reflect propeller's working characteristics, so the propeller's features can be extracted directly from these line components. The other way to extract the features is from demodulated line component.So there are two ways to extract the features, the direct way and the demodulation way. Detection performance of the line component in background-noise is discussed in this paper. The signal level is defined as the difference between the PDF's (Probability Density Function) mean of the peak of the line component and PDF's mean of the background-noise. In the direct way the signal level of the line component is proportional to the signal to noise ratio (S/N). In the demodulation way the signal level of demodulated line component reduces with S/N descent, but the signal level reduces slowly if S/N is high and very fast if S/N is low.