Signal processing by means of noise

Noisy integrate-and-fire neurons encode periodic signals best at a preferred noise-frequency combination. We show that the preferred frequency, optimum noise and optimum signal-to-noise ratio are linearly related to the signal AC amplitude with the signal DC amplitude serving as tuning parameter. We argue that this may facilitate selective signal processing in neuronal networks. (C) 2001 Elsevier Science B.V. All rights reserved.