A DEMON computation method based on variational mode decomposition with multifeature weighted reconstruction

Detection of Envelope Modulation of Noise (DEMON) is a primary technique utilized in sonar signal processing. The classic DEMON depends on prior knowledge to determine the demodulation band, resulting in limited adaptability to interference and exhibiting inadequate line spectral strength and stability. An enhanced DEMON calculating methodology is proposed. The proposed method firstly decomposes the signal over different time scales using variational mode decomposition (VMD). The time-domain Pearson correlation coefficient and spectral kurtosis are computed, and the intrinsic modal functions (IMFs) are objectively assessed and weighted to facilitate reconstruction for interference suppression and line spectrum enhancement. Simulation and experimental data indicate that, in comparison to the classic DEMON, the proposed DEMON, which utilizes VMD with multi-feature weighted reconstruction, exhibits enhanced line spectral stability and signal-to-noise ratio (SNR). At low SNR, more pronounced harmonic structures are discernible, enhancing target recognition and classification.