一类非线性网络系统中噪声改善信号的相关性

Noise-improved Signal Correlation in a Nonlinear Network System

基于相关系数研究了在一类非线性神经网络系统中加性和乘性噪声作用下的阈上随机共振现象。仅在加性噪声或者乘性噪声的作用下,对每一个固定的系统阈值,加性噪声下的阈上随机共振比乘性噪声下的阈上随机共振更容易发生,且相关系数所达到的峰值也比在乘性噪声下的峰值大,这说明加性噪声更有利于改善信号的相关性。系统阈值的增加会降低阈上随机共振的功效;而阈值单元数目的增多,会提高阈上随机共振的功效。加性和乘性噪声共同作用下的阈上随机共振现象同样存在,对系统阈值进行恰当选取和增加系统阈值单元数目使得阈上随机共振现象更加明显;给定乘性噪声而改变加性噪声比固定加性噪声而改变乘性噪声阈上随机共振更容易发生,且功效更好。

Based on correlation coefficient the Suprathreshold Stochastic Resonance （ SSR） phenomenon under additive and multiplicative noise in a nonlinear neural network system is studied. When the system is only affected by one kind of noise,e. g. the additive noise or multiplicative noise,under different system threshold,additive noise is easier for SSR phenomenon to occur than multiplicative noise. For a fixed threshold,the additive noise is superior to multiplicative noise for signal transmission since the resonance peak value under the ad-ditive noise is larger than that under the multiplicative noise. The increase of network threshold cuts down the efficiency of SSR,while the increase of unit number plays a positive role in the emergence and effect of SSR. The composing noise mixed by additive noise and multi-plicative noise can also arouse SSR in such a system. A reasonable threshold value and a larger unit number would lead to a more signifi-cant SSR. Fixing multiplicative noise and changing additive noise,the SSR phenomenon is more likely to happen and gets a better effect than under the contrary operation. It is an extension of noise benefits in signal processing.