In the field of underwater acoustic signal processing, the ship radiated noise contains a large amount of ship information, which is of great significance to the ship identification. The traditional method relies too much on the operator and prior knowledge, which seriously reduces the efficiency and accuracy of the ship radiated noise identification. This paper presented a novel ship radiated noise feature extraction method based on compression sensing and center frequency. Firstly, to compression sensing of the ship radiated noise, enhance its line spectrum energy. Then, the ship radiated noise is decomposed by empirical mode decomposition to obtain multiple intrinsic mode function, calculate the mutual information entropy of adjacent intrinsic mode function to determine the key parameter K of the variational mode decomposition. Finally, perform variational model of ship radiated noise based on K, extract the center frequency of maximum energy intrinsic mode function as the ship radiated noise recognition feature. Experimental results show that the proposed feature extraction method can classify ship radiated noise quickly and effectively, and reduce the dependence on operators and prior knowledge.

Accurate underwater target detection and recognition in complex marine environments has always been a challenge. There is a lot of information in underwater target radiation noise that is important for underwater target recognition. However, the traditional underwater target radiation noise process is inefficient and inaccurate, severely limiting underwater target recognition. This paper proposed a new method for underwater target recognition based on compressed sensing multiscale entropy. For starters, compressing a signal improves its signal-to-noise ratio and broadens its linear spectrum. The multiscale sample entropy for the signal is then calculated after it has been denoised, and the most separated sample entropy is chosen by comparing the different scales of sample entropy to achieve effective underwater target radiation noise recognition. The experimental results show that the feature extraction method proposed in the paper can classify underwater target radiation noise quickly and effectively, improving recognition efficiency.