The Research Of Parameter-tuning SR (PSR) Via ANN

Stochastic Resonance is a new and important embranchment in the field of nonlinear science. This phenomenon has attracted much attention for the past two decades, and it has been observed that SR can occur in a wide variety of systems. With the development of its research, it has been applied in signal processing. In the application of weak signal detection, SR is generated by adjust noise intensity at first, and PSR is suggested later. PSR means adjusting system parameter to generate stochastic resonance, but putting PSR into hardware practice is not ideal, and PSR is not fit for short time processing.Consider of the performances of NN which are high parallel, nicer fault-tolerant and associate memory, powerful adaptive and self-learned ability, this thesis researches on the needed system parameters when bistable system reach resonance at different signal intensity and noise intensity, and studies the nonlinear dynamic system identification with fluctuant parameters and using multi-functional link artificial NNs. Finally, the network is used to realize PSR. In the first part of this thesis, the history, the development and the application in signal processing of SR are introduced briefly. And the main work of this paper is described. Chapter 2 introduces the theory and the simulation of SR.In Chapter 3, we research on the needed system parameters when bistable system reaches resonance at different signal intensity and noise intensity.In Chapter 4, we study fluctuant parameter nonlinear dynamic system identification using multi-functional link artificial neural networks. And the trained network is used to realize PSR. Finally, we draw some conclusions of the work of this paper and put forward the possible future works..