Research on the line spectrum denoising detection based on multi-scale feature autoencoder

Abstract Line spectrum detection is an important research direction in the field of underwater acoustic target detection. The relevant features can be obtained by using line signal power spectrogram, but radiated noise susceptible to influence external noise, which degrades quality and thus affects complete extraction spectrum. A denoising method proposed this paper based on deep learning network architecture classical filter extract from spectrogram with strong background while maintaining continuity improving efficiency simulated used train multi-scale feature autoencoder remove different scale interference part then pass morphological attribute further keep continuity. experimental analysis measured fishing vessel shows that algorithm effectively interference, improve signal-to-noise ratio quality, realize effective