Research On Signal Reconstruction Method Of Weighted Super-Threshold Stochastic Resonance

Suprathreshold stochastic resonance(SSR)is a distinct form of stochastic resonance,which occurs in multilevel parallel threshold arrays.SSR is an important extension of stochastic resonance with potential applications in information processing and biomedical engineering fields,since it operates with signals of arbitrary magnitude,not restricted to weak or sub-threshold signals.The generic model of SSR can be described in terms of stochastic quantizer.In this dissertation,we investigate the decoding scheme and performance of a quantized signal in the SSR model.The main contents of this dissertation can be summarized as follows:1.A new decoding scheme is proposed,under the name of optimal weighted decoding for stochastic quantifying signal,based on the generic SSR model,a summing array of threshold subsystems is modeled,and optimal weighted coefficients and the expression of the reconstructed output are all derived.The mean square error(MSE)distortions of this decoding scheme for three examples of threshold settings,i.e.,identical,unique and group thresholds,are analyzed in detail.The obtained results show that the MSE distortion is the lowest for the case of group thresholds.In this situation,the MSE distortion of optimal weighted decoding is improved over that of Wiener linear decoding.2.A multigroup weighted decoding scheme is proposed based on optimal weighted decoding for stochastic quantifying signal.Moreover the weighted decoding scheme in a summing array of threshold subsystems is extended to arbitrary nonlinear arrays.We especially apply this multigroup weighted decoding scheme to a parallel array of saturating sensors.MSE distortion comparison between the optimal weighted decoding and Wiener linear decoding are investigated in detail.The results show that,for the case of more than two groups with equal interval shifted parameters,the optimal weighted decoding scheme is superior to Wiener linear decoding.Moreover,the MSE distortions significantly decrease as the group size increases,and achievethe minimum when group sizes are equal to array sizes.In addition,for the case of various slope,the MSE distortion of multigroup weighted decoding is also improved over Wiener linear decoding except very small slope.3.An adaptive weighted decoding scheme is proposed by utilizing the adaptive filtering theory.Thus this weighted decoding scheme is extended to more general input characteristics.We apply the decoding scheme to a parallel array of threshold elements and investigate the decoding performance for inputs with stationary,non-stationary characteristics under white Gaussian noise and colored noise circumstances.The results exhibit that,requiring little or no a priori knowledge of the signal,the adaptive weighted decoding can be applicable to not only the simple case of stationary signal,but also the non-stationary inputs and the colored noise situations.In addition,the SSR model can also be considered as the channel of information transmission.The noise-enhanced information transmission with generalized Gaussian signal is investigated.The effect of the exponent parameter on maximum information transmission is discussed.It is found that the exponent parameter of the generalized Gaussian noise distribution has a crucial effect on the position or the maximum value of the mutual information.The weighted decoding schemes on the signal reconstruction in SSR were explored.These schemes may help promote the SSR theory develop and spread in engineering applications.