Signal Selection Based on Stochastic Resonance

Noise aids the encoding of continuous signals into pulse sequences by way of stochastic resonance and endows the encoding device with a preferred frequency. We study encoding by a threshold device based on the Ornstein-Uhlenbeck process, equivalent to the leaky integrate-and-fire neuron model. Preferred frequency, optimum noise intensity, and optimum signal-to-noise ratio are shown to be linearly related to the AC amplitude of the input signal. The DC component of the input tunes the device either into transmission (preferred frequency nearly independent of signal amplitude) or selection mode (frequency rising with amplitude). We argue that this behavior may facilitate selective signal processing in neurons.