Noise-improved signal correlation in an array of autoregressive models of order one

摘要

This paper discusses noise-improved signal correlation through an array of autoregressive models of order one [AR (1)]. In a single model, when the input is a square wave or a cosine with a higher frequency, stochastic resonance (SR) exists. When the input is a cosine with a lower frequency, SR and suprathreshold stochastic resonance (SSR) both exist. SSR can also exist for a cosine with a higher frequency ora square wave through an array of AR (1) models. The efficacy of SR and SSR increases as the number of AR (1) models is raised or as the threshold is lowered in the array. There is a range of values of noise standard deviation where the correlation coefficient between the input and output signals is greater than that of the input signal and the noisy signal.