Optimal quantization and suprathreshold stochastic resonance

Abstract

It is shown that Suprathreshold Stochastic Resonance (SSR) iseffectively a way of using noise to perform quantization or lossysignal compression with a population of identical threshold-baseddevices. Quantization of an analog signal is a fundamentalrequirement for its efficient storage or compression in a digitalsystem. This process will always result in a loss of quality,known as distortion, in a reproduction of the original signal. Thedistortion can be decreased by increasing the number of statesavailable for encoding the signal (measured by the rate, or mutualinformation). Hence, designing a quantizer requires a tradeoffbetween distortion and rate. Quantization theory has recently beenapplied to the analysis of neural coding and here we examine thepossibility that SSR is a possible mechanism used by populationsof sensory neurons to quantize signals. In particular, we analyzethe rate-distortion performance of SSR for a range of input SNR'sand show that both the optimal distortion and optimal rate occursfor an input SNR of about 0 dB, which is a biologically plausiblesituation. Furthermore, we relax the constraint that allthresholds are identical, and find the optimal threshold valuesfor a range of input SNRs. We find that for sufficiently smallinput SNRs, the optimal quantizer is one in which all thresholdsare identical, that is, the SSR situation is optimal in this case.