Study On Weak Signal Detection Based On Stochastic Resonance Theory

Weak signal detection is an urgent crucial technology in deep space communication,long-range target detection and electromagnetic information leakage detection.In weak signal detection,most researchers improve the performance by suppressing noise components.Conventional methods of linear processing suppress noise and weaken signal energy simultaneously,which are not suitable for complex environments with robust background noise.However,non-linear Stochastic Resonance(SR)could enhance the signal energy without eliminating noise interference or even add a certain intensity of noise components.Therefore,this dissertation focuses on solving the complex problems in weak signal detection based on stochastic resonance theory,centering on the defect of inherent saturation,the influence of system parameters,detection of multi-target signals,limitations of short memory,environment of non-Gaussian noise,etc.And the corresponding weak signal detection algorithms based on stochastic resonance theory are proposed.This dissertation revolves around the four aspects:1.An unsaturated bistable stochastic resonance signal detection algorithm based on the piecewise potential functionThe inherent output saturation defect in the classical bistable stochastic resonance algorithm limits the amplitude of the output signal.This dissertation analyzes the cause of these limits and their adverse effects on weak signal detection.An exponential function is used to construct a piecewise potential role to avoid the influence of higher-order terms.The Unsaturated Bistable Stochastic Resonance(UBSR)algorithm is proposed,whose Kramers escape rate and output signal-to-noise ratio are theoretically deduced.Compared with the classical bistable stochastic resonance algorithm,the proposed algorithm could avoid output saturation effectively and achieve better and faster detection performance under intensive background noise.2.An adaptive unsaturated bistable stochastic resonance signal detection algorithm based on genetic algorithmIt is difficult for the input terminal under intensive background noise to only adjust the intensity of the excitation noise to ensure the synergy of the non-linear system,periodic signal and stochastic noise.This dissertation adds an adaptive algorithm to further optimize the relevant parameters of the stochastic resonance algorithm for promoting stochastic resonance based on the improved piecewise potential function.In addition,the fitness function is used as a criterion for judging the detection performance among adaptive algorithms,which significantly affects the direction of optimization and the speed of convergence.However,it is difficult for the existing fitness function to measure signal periodicity and frequency accuracy simultaneously.Therefore,a fitness function,which is more suitable for weak periodic signals,is constructed to judge the occurrence of stochastic resonance phenomenon intuitively.An Adaptive Unsaturated Bistable Stochastic Resonance(AUBSR)algorithm is proposed,combined with a genetic algorithm for global optimization.Experiments show that the proposed algorithm effectively enhances the detection performance of weak periodic signal under robust Gaussian noise.3.An adaptive unsaturated bistable stochastic resonance signal detection algorithm based on empirical mode decompositionThe stochastic resonance algorithm is able to detect single weak signal with small parameter effectively,whose flaw lies in detecting multi-target signals under intensive background noise.The target components with high frequency will be filtered out possibly,which may cause the omission of signal detection.This dissertation introduces Empirical Mode Decomposition(EMD)that can quickly decompose multi-frequency components into the stochastic resonance.According to the fitness function value,the components near the target frequency are screened,which avoids mode aliasing caused by the direct use of EMD components.Meanwhile,an Adaptive Unsaturated Bistable Stochastic Resonance based on Empirical Mode Decomposition(EMD-AUBSR)is proposed to handle large parameter signal detection,which utilizes re-sampling to expand the frequency range of signals to be measured.Experiments show that the proposed algorithm realizes multi-target signals detection under intensive Gaussian noise.4.An adaptive unsaturated bistable stochastic resonance signal detection algorithm based on time-delayed feedbackSince the stochastic resonance algorithm is a short-time memory system whose current output sequence only relates to the previous output sequence.Thus,the algorithm is inadaptable under complex noise.This dissertation further constructs the multi-target signals detection algorithm under non-Gaussian background noise based on the signal detection algorithms under robust Gaussian noise.Taking an example for the non-Gaussian Lévy noise environment,this dissertation proposes Adaptive Unsaturated Bistable Stochastic Resonance based on Time-delayed Feedback(TF-AUBSR).The algorithm takes time-delay feedback items as the historical information of periodic signals to compensate for the deficiency of the original algorithm.In this way,it transforms the stochastic resonance algorithm into a stochastic system with a long-time memory.Experiments show that the proposed algorithm could effectively detect multi-target signals under complex noise.