Noise can assist neurons in the detection of weak signals via a mechanism known as stochastic resonance (SR). We demonstrate experimentally that SR-type effects can be obtained in rat sensory neurons with white noise, $1/f$ noise, or $1/f^2$ noise. For low-frequency input noise, we show that the optimal noise intensity is the lowest and the output signal-to-noise ratio the highest for conventional white noise. We also show that under certain circumstances, $1/f$ noise can be better than white noise for enhancing the response of a neuron to a weak signal. We present a theory to account for these results and discuss the biological implications of $1/f$ noise.

• Received 2 September 1998

DOI:https://doi.org/10.1103/PhysRevLett.82.2402