Deep stochastic resonance array and its application in enhancing underwater weak signals

Enhancing underwater weak signals, such as ship radiation in high background noise environments, is crucial for remote target detection. Stochastic resonance (SR) and its complex systems, which utilize noise to amplify weak signals, have shown significant promise in underwater signal enhancement. While parallel stochastic resonance arrays (PSRAs) effectively enhance weak signals through the collective effect of multiple subsystems, traditional PSRA methods fail when noise intensity exceeds the resonance region and are limited to single-layer arrays, restricting their potential. To address these limitations, we propose the Deep Stochastic Resonance Array (DSRA) method. DSRA overcomes the failure of traditional PSRAs in high-noise environments by selecting appropriate nonlinear systems with suitable resonance regions and injecting an optimal amount of additional noise to disrupt noise correlation between parallel subsystems. DSRA employs a cascading multi-layer parallel array structure, leveraging the reduced correlation between residual output noise and the accompanying input noise after additional noise injection. By feeding the original noisy signal into each layer, DSRA ensures complete target information in each layer, achieving progressive signal enhancement. Simulation comparisons show that DSRA exhibits superior band-pass filtering characteristics, with an SNR gain exceeding 23 dB compared to PSRA and other methods. DSRA also demonstrates an anti-noise performance improvement of over 18 dB relative to existing methods. Stability tests indicate that DSRA outperforms SR and cascaded stochastic resonance (CSR) by 30% at −28 dB and surpasses PSRA and similar methods by more than 15%. Application verification experiments further confirm the practical applicability and superior performance of DSRA in underwater acoustic signal processing. This study highlights the potential of complex SR systems for significant weak signal enhancement.