Stochastic quantization transfer functions for high resolution signal estimation

Abstract

Scalar signal quantization is commonly based on a deterministic definition of nonlinear transfer functions. However, in the more general case of stochastic quantization a non-deterministic approach is required. This paper presents an analysis of general stochastic quantization properties derived from a model incorporating a regular deterministic quantizer. It is shown how stochastic quantization transfer functions providing constant mean square error independent of the actual probability density distribution of the signal being quantized can be derived. Simulation results illustrating performance and limits of the proposed quantizer are given.