Stochastic Resonance — Subthreshold Signal Detection

Signal & Noise

Signal amplitude A 0.7 Threshold θ 1.0 Noise σ 0.5 Frequency f 0.05

SNR Readout

Signal SNR—

SNR

Firing rate—

Subthreshold?—

Optimal σ*—

About

Stochastic resonance (Benzi 1981, Gammaitoni 1998): adding a small amount of noise to a subthreshold signal can improve detection. The threshold neuron fires when signal + noise exceeds θ. Too little noise → no firing; too much → random firing drowns the signal. An optimal noise level maximizes SNR. The SNR-vs-σ plot (bottom) shows the resonance peak.