Underwater acoustic weak signal detection based on noise reduction and reconstruction of variational modal components

Aiming at the problem that the performance of the energy detection methods for non-cooperative underwater acoustic signal detection decreases under low signal-to-noise ratio, a signal detection method combining variational modal decomposition and wavelet transform to reduce the noise and reconstruct signal is proposed. In this method, the ratio of approximate entropy and cross-correlation coefficient of each intrinsic mode function obtained by signal decomposition is used as the component classification parameter. The obtained components are divided into signal components, noisy signal components and noise components. The noisy signal components are denoised by the second generation wavelet transform and then combined with the signal components to form the reconstructed signal. Finally, the reconstructed signal will be detected. Numerical simulation results show that the proposed method can reduce the noise of CW and LFM signals adaptively without prior information, and can improve the detection probability under low false alarm probability. When the SNR is below 0 dB, the SNR of CW signal can be improved by about 12 dB after noise reduction, and the SNR of LFM signal can be improved by about 8−9 dB. The effectiveness of the proposed method was verified by lake test data. When the false alarm probability was 0.1, the detection probability of the proposed method was increased to above 0.9.