Research On Stochastic Resonance Theory And Signal Enhancement Methods Based On Biological Neurons

In order to enhance weak signals by using noise and stochastic resonance,based on the Hodgkin-Huxley(H-H)neuron,Fitzhugh-Nagumo(FHN)neuron and Izhikevich neuron models,this paper introduces three different types of noise into the input of neurons,namely Gaussian white noise,Lévy noise and Colored noise,and studies the phenomena of subthreshold stochastic resonance,Suprathreshold stochastic resonance(SSR)and double stochastic resonance in single H-H neurons,FHN neurons and Izhikevich neurons and the weak signals were amplified successfully by three kinds of neurons.First of all,the responses of three kinds of neurons to different frequency,amplitude and constant offset stimuli were studied,and the membrane potential and its spectrum were given.For H-H neurons,the Mutual information rate(MIR)between the output signal and the input signal of the neuron is taken as the evaluation index.The research results show that H-H neurons can amplify the input signal more effectively under certain intensity of noise stimulation,and exist in a non-zero noise intensity range.MIR will have a maximum value at this intensity,which proves the existence of stochastic resonance and SSR phenomena under the threshold.It is also observed that neurons have frequency selectivity,and there are also differences in transmission performance for input signals with different amplitudes and constant offsets.Finally,it is observed that the adaptability of the stochastic resonance characteristics of the three kinds of noise induced neurons is different.The stochastic resonance of H-H neurons induced by Gaussian white noise can better amplify weak signals.For FHN neurons,it is observed that the MIR of neurons induced by Gaussian white noise has double peaks through the 3D and plane diagrams of MIR,indicating the appearance of double stochastic resonance phenomenon.In addition,it is observed that the neural model induced by Gaussian white noise and colored noise has almost the same transmission performance for input signals,while the neural model induced by Lévy noise is unstable.It is believed that the Stochastic resonance of FHN neurons induced by Gaussian white noise can better amplify weak signals.For Izhikevich neurons,the most representative inhibitory Fast spiking(FS)state model and Chattering(CH)state model among Izhikevich neurons were selected.Through the 3D and plane diagrams of MIR,we can see that MIR has unimodal under different conditions.In addition,it is found that the bright region of FS state model is narrow,while that of CH state model is wide.The study also found that the MIR is almost the same under low frequency and high frequency input signals.However,with the increase of constant offset stimulus and pulse height,the MIR decreases,and the bright areas of FS and CH state neuron models will shift to both sides respectively.In general,the stochastic resonance characteristics of the CH state neuron model induced by Gaussian white noise are better than those of the FS state neuron model,followed by Lévy noise and Colored noise.More importantly,the research results of this paper will help to further study the role of these three kinds of noise and more noise in the stochastic resonance of more complex biological neural networks,and further promote its development in the fields of engineering,medicine,bionics and artificial intelligence.