Stochastic resonance (SR) can be used to recover a weak linear frequency modulated (LFM, chirp) signal with unknown parameters from strong noise by guiding an adaptive bistable SR system with a fractional-Fourier-transform-based performance index and a cloud model genetic algorithm (CMGA). The method was demonstrated in 2020 only through numerical simulations with Gaussian white noise and a small parameter set. The open problem is to establish the practical validity domain of SR-based enhancement of unknown LFM signals: behavior under non-Gaussian and colored noise and at extremely low signal-to-noise ratios, systematic (rather than empirical) selection of the adaptive-SR control coefficients, robustness when the instantaneous frequency deviates from an exact linear chirp, experimental validation on real measured signals such as radar echoes or machinery vibrations, and the computational cost of the fractional Fourier transform plus CMGA pipeline for large data sets.

Background

Stochastic resonance (SR) is a nonlinear phenomenon in which a weak input signal, added noise, and a nonlinear system cooperate: noise energy is partially converted into signal energy, so that a signal too weak to detect can be amplified when the noise intensity and system parameters are matched to a resonance-like operating point. Classically SR is realized in a bistable system, i.e. a system whose potential has two wells separated by a barrier, such as the overdamped particle dynamics x˙=ax−bx3+s(t)+n(t)\dot{x}=ax-bx^{3}+s(t)+n(t) with a,b>0a,b>0; the input s(t)s(t) plus noise n(t)n(t) drives transitions between the wells. For periodic inputs, standard performance indices exist: output signal-to-noise ratio (SNR), spectral amplification factor, and the cross-correlation coefficient between input and output. A linear frequency modulated (LFM) signal, also called a chirp, is a signal of constant amplitude whose instantaneous frequency varies linearly in time, s(t)=Acos⁡(2π(f0t+12kt2))s(t)=A\cos(2\pi(f_{0}t+\tfrac{1}{2}kt^{2})); LFM signals are ubiquitous in radar, sonar, and covert communications because of their large time-bandwidth product. Because a chirp is aperiodic and sweeps through frequencies, the classical periodic-input indices lose their meaning, which blocked the application of SR to unknown LFM signals. A series of works addressed this: SR for aperiodic LFM signals with random frequency fluctuation (Lin, Wang and Lv 2017), adaptive piecewise re-scaled SR excited by LFM signals (Wu, Wang, Yang, Huang and Sanjuán 2020), where a re-scaling compresses the time axis so a high-frequency chirp behaves adiabatically, and LFM-induced aperiodic resonance with a modified cross-correlation index (Zhai, Yang, Zhang and Liu 2020). The pivotal step for this problem is the paper by Wu, Yang, Sanjuán and Liu (Chaos 30, 043128, 2020): they proposed to measure SR performance for an unknown LFM signal by a fractional signal-to-noise ratio (FSNR), an SNR-like index computed in the domain of the fractional Fourier transform (FRFT), a one-parameter family of integral transforms of which the ordinary Fourier transform is a special case and in which a chirp of the right rate concentrates into a sharp peak, exactly as a sinusoid concentrates in the ordinary frequency domain. Maximizing this FSNR over the SR system parameters lets the system recover an unknown LFM signal directly from a raw noisy record, and the parameter search is automated by a cloud model based genetic algorithm (CMGA), a genetic algorithm whose crossover and mutation strengths are governed by cloud-model uncertainty rules controlled by a set of numerical coefficients. The demonstration, however, covered only Gaussian white noise and a limited set of chirp parameters, with coefficients tuned empirically, and it left open how the method behaves under realistic interference, model error, real instruments, and large data volumes. That unvalidated envelope is the subject of this problem, and it matters for practitioners in radar, sonar, and rotating-machinery monitoring who consider SR as a weak-signal-enhancement stage in real receivers.

Problem Statement

Establish the practical validity domain of stochastic-resonance-based recovery of unknown-parameter LFM signals in strong noise, as introduced by Wu, Yang, Sanjuán and Liu (2020) with the FSNR index and CMGA-tuned adaptive SR, beyond its original Gaussian-white-noise numerical demonstration. Directions in which this validity domain remains open beyond that demonstration are the following: (1) the method's detection and recovery performance when the noise is non-Gaussian or colored, and its operating limit at extremely low input signal-to-noise ratios, quantified through the FSNR/output behavior of the SR system rather than anecdotal examples; (2) systematic, non-empirical guidance for selecting the adaptive-SR and CMGA control parameters, supported by a sensitivity analysis over the tuning coefficients so that a practitioner can choose them without trial-and-error; (3) the degradation of the FSNR index and of signal recovery when the hidden signal's instantaneous frequency deviates from an exact linear chirp, i.e. the robustness to LFM model mismatch; (4) whether the enhancement works on real measured strong-noise records such as radar or sonar echoes or rotating-machinery vibration data containing weak chirp components, not only on synthetic data, with recovery evaluated against an independently established reference for that record; and (5) the computational cost and scalability of the FRFT plus CMGA pipeline for large data sets. Together these directions delimit the validated operating envelope of SR enhancement for unknown LFM signals that the problem asks to characterize; the envelope remains unestablished while any one of the directions is unresolved. That envelope is characterized, for each direction, by the range actually covered (which noise types, SNR values, chirp deviations, records, or data sizes were tested) together with the limit or breakdown observed; any untested region of it remains open rather than resolved. Relative terms such as 'extremely low SNR' and 'large data sets' are to be interpreted against the conventions of the weak-signal-detection literature in which the source is situated, not by invented fixed thresholds.

References

  1. Chengjin Wu, Jianhua Yang, Miguel A. F. Sanjuán, Houguang Liu, 'Stochastic resonance induced by an unknown linear frequency modulated signal in a strong noise background', Chaos 30, 043128 (2020). DOI: 10.1063/5.0002134, https://doi.org/10.1063/5.0002134
  2. Yuwen Zhai, Jianhua Yang, Shuai Zhang, Houguang Liu, 'Linear frequency modulated signal induced aperiodic resonance', Physica Scripta 95, 065213 (2020). DOI: 10.1088/1402-4896/ab825a, https://doi.org/10.1088/1402-4896/ab825a
  3. Chengjin Wu, Zhiqiang Wang, Jianhua Yang, Dawen Huang, Miguel A. F. Sanjuán, 'Adaptive piecewise re-scaled stochastic resonance excited by the LFM signal', The European Physical Journal Plus 135, 130 (2020). DOI: 10.1140/epjp/s13360-020-00195-0, https://doi.org/10.1140/epjp/s13360-020-00195-0
  4. Lifeng Lin, Huiqi Wang, Wangyong Lv, 'Stochastic resonance system with linear random frequency fluctuation for aperiodic LFM signal', Nonlinear Dynamics 88, 1361-1371 (2017). DOI: 10.1007/s11071-016-3315-y, https://doi.org/10.1007/s11071-016-3315-y
  5. Chen Yang, Jianhua Yang, Dengji Zhou, Shuai Zhang, Grzegorz Litak, 'Adaptive stochastic resonance in bistable system driven by noisy NLFM signal: phenomenon and application', Philosophical Transactions of the Royal Society A 379, 20200239 (2021). DOI: 10.1098/rsta.2020.0239, https://doi.org/10.1098/rsta.2020.0239
  6. Jin Liu, Zan Li, Yue Gao, Qiguang Miao, Li Yang, Danyang Wang, Peihan Qi, 'Adaptive stochastic resonance enhanced weak linear frequency modulated signal perception in low signal-to-noise ratio environments', Measurement Science and Technology 36, 016102 (2024). DOI: 10.1088/1361-6501/ad8255, https://doi.org/10.1088/1361-6501/ad8255
  7. Jianhua Yang, Miguel A. F. Sanjuan, 'Aperiodic Stochastic Resonance Caused by Frequency-Modulated Signal and Noise', chapter in 'Aperiodic Resonances - Theory and Applications', Springer Nature Singapore (2025), pp. 79-102. DOI: 10.1007/978-981-96-6819-9_5, https://doi.org/10.1007/978-981-96-6819-9_5
  8. Yang Liu, Zhenhua Xiong, 'A generalized stochastic resonance based instantaneous frequency estimation method under low SNR', Mechanical Systems and Signal Processing 164, 108269 (2022). DOI: 10.1016/j.ymssp.2021.108269, https://doi.org/10.1016/j.ymssp.2021.108269
  9. Gang Zhang, Yichen Shu, Tianqi Zhang, 'Piecewise unsaturated multi-stable stochastic resonance under trichotomous noise and its application in bearing fault diagnosis', Results in Physics 30, 104907 (2021). DOI: 10.1016/j.rinp.2021.104907, https://doi.org/10.1016/j.rinp.2021.104907
  10. Gang Zhang, Yujie Zeng, Tianqi Zhang, 'Second-order coupled tristable stochastic resonance and its application in bearing fault detection under different noises', Nonlinear Dynamics 111, 8987-9009 (2023). DOI: 10.1007/s11071-023-08303-2, https://doi.org/10.1007/s11071-023-08303-2
  11. Baofeng Li, Zhongqiu Wang, Jianhua Yang, Tao Gong, Miguel A. F. Sanjuán, 'Stochastic resonance as a tool for smoothing time-frequency ridges in nonstationary signals', Europhysics Letters 151, 61002 (2025). DOI: 10.1209/0295-5075/ae0366, https://doi.org/10.1209/0295-5075/ae0366