Abstract:
In order to obtain effective target information in the measured ship radiated noise signal and improve the separability of target signals under low signal to noise ratio (SNR) condition, a feature extraction method for ship radiated noise signal based on variational mode decomposition (VMD) and resonance-based sparsity signal decomposition (RSSD) is proposed in this paper. Firstly, based on the fact that the ship radiated noise signal is periodic and the noise is random, the VMD autocorrelation analysis method is used to reconstruct the signal and mainly eliminate the out-of-band noise components. Then, based on the different resonance properties of the signal, RSSD algorithm is used to further filter the in-band noise and transient interference, and realize the extraction of periodic oscillation components in the signal. Finally, the waveform structure features of the signal are extracted and used for target classification and recognition. The analysis results of simulation signal and the measured signal analysis show that the method can filter out the out-of-band and in-band noise well and enhance the inherent narrow-band characteristics of ship radiated noise signal. The experimental results of multi-class ship target classification show that this method can effectively improve the separability of low SNR signals and improve the performance of target recognition.