Propeller Shaft Frequency Extraction Based on the Improved Cyclic Modulation Spectrum

Propeller noise is one of the main radiation sources of underwater targets. It contains substantial physical information for feature extraction and recognition. Most of the extraction methods of propeller shaft frequency based on the Detection of Envelope Modulation on Noise (DEMON), which have the disadvantages of large computation cost and low reliability, some even require the prior information of target. In this paper, we propose to combine wavelet and Hilbert- Huang transform (HHT) to suppress background noise, and a shaft frequency method based on Multiple Signal Classification (MUSIC) is used to reduce the computation of Cyclic Modulation Spectrum (CMS). Here, we have two contributions: firstly, HHT-based improved envelope is exploited for more thorough filtering of low-band continuum noise. Secondly, the searching method of shaft frequency and harmonics improve the searching accuracy in both multi-target and single-target detection.