Research On Stochastic Resonance Based On RLGA And Its Application To Underwater Acoustic Signal Detection

We are continuing to invest in the field of underwater communications research,one of which is the detection of signals in communication.Overwhelmed by a large amount of noise caused by ships,marine life and man or interference of refraction and frequency selective weakness,the signal often becomes weak signal through the underwater acoustic channel.The detection technology of underwater acoustic signal has important significance and application value both in theoretical research and in practical engineering and even national defense construction.Therefore,the research of weak signal detection method is necessary for the progress and development of underwater acoustic communication.Some traditionalmethods such as matched filtering,coherent detection,time-frequency analysis,etc.detect signal through suppressing noise,which although can achieve very good results under a certain condition,but damage the useful signal in the frequency structure or the time domain amplitude while suppressing the noise.The difference between stochastic resonance and conventional detection is that it is not intended to filter out noise,but to maximize noise,and to convert noise energy into signal energy through a nonlinear system,thereby enhancing the effect of weak signals.This provides a new idea for signal detection in low-signal-to-noise environments in underwater acoustic channels.The general theoretical approach of stochastic resonance is to gradually increase the noise intensity of the known signal to achieve the starting conditions.However,in practical applications of underwater acoustic communication,the intensity of signal and noise are unknown,whichrequires a system to be able to adjust the parameters to achieve stochastic resonance.In this paper,the adaptive adjustment of the system parameters is realized by using the working mechanism of reinforcement learning combined with genetic algorithm:Genetic algorithm provides a framework of genetic space for reinforcement learning,and reinforcement learning replaces the cross,reorganization and mutation of genetic algorithms.The algorithm uses the output signal to noise ratio as the fitness evaluation index of the system of stochastic resonance.Therefore,this paper also comparesdifferent SNR estimation algorithms and selects the one that is most suitable for the environment.In addition,the ocean noise is simulated and superimposed on the source signal to simulate the underwater acousticnoisy signal.It is convenient to study the underwater acoustic communication signal without the experimental condition.In this paper,a series of MATLAB simulation experiments show that the proposed stochastic resonance method based on reinforcement learning and genetic algorithm is feasible for the detection of underwater acoustic signals,and the signal-to-noise ratio of the signal can be enhanced.