Research On The Application Of Array Neuron Model Based On Stochastic Resonance

The emergence of stochastic resonance subverts people's cognition of noise.Generally,noise is considered as the irregular information generated in the transmission process of information,which interferes with our extraction and utilization of original information.The existence of noise seriously affects the quality of data.The stochastic resonance effect provides a new way to deal with noise.Stochastic resonance effect is common in neurons,but there are few studies on stochastic resonance in neurons.In this paper,stochastic resonance effect is combined with neuron system.Aiming at the problem of weak signal enhancement and low peak signal-to-noise ratio image restoration,an array neuron model based on stochastic resonance is proposed.The main work and research results are as follows:(1)In this paper,an array saturated synaptic neuron model based on stochastic resonance(SR)is established.Different sizes of series and parallel saturated synaptic neuron models are used to enhance weak periodic and aperiodic signals.The effects of noise intensity and array size on weak signal enhancement were studied by changing the noise intensity or array size,and the advantages and disadvantages of the same scale of series and parallel arrays of saturated synaptic neurons on weak signal enhancement were analyzed.The experimental results show that the parallel neuron model has better effect on weak signal enhancement and better restoration of signal details.(2)In this paper,a parallel Fitzhugh Nagumo(FHN)neuron model based on stochastic resonance is established for low peak signal to noise ratio(PSNR)gray image restoration.By changing the size and noise intensity of FHN neuron array,the influence of array size and noise intensity on image restoration results is analyzed.Then,the peak signal-to-noise ratio,bit error rate and structural similarity are used as evaluation indexes to objectively judge the restoration effect of low PSNR image.The parallel array method of FHN neurons is compared with the traditional filtering method and two-dimensional stochastic resonance method.Experimental results show that this method can greatly improve the PSNR value of low PSNR gray image,and the visual effect of restored image is excellent.(3)A new gray image preprocessing method is proposed.In this method,the original two-dimensional image is first converted into a one-dimensional signal through a row scan,and then each gray value of the one-dimensional signal is converted into an eight-bit binary.Then the binary sequence is processed by the array neuron system after polarity conversion and amplitude modulation,and the signal processed by the array neuron system is decoded and anti-scanning processed,and finally the restored grayscale image is obtained.The restored image information obtained by this method is more complete than the general method,and has better visual effect.