Distributed estimation based on quantized data

Since standard statistical estimation methods are built on the models that treat numerical data as continuous variables, they can be inappropriate and misleading when quantization process is involved in estimation. In this paper, we propose novel distributed estimation algorithms based on the Maximum Likelihood (ML) method. Motivated by the observation that each quantized measurement corresponds to a region with which the parameter to be estimated is associated, we develop algorithms that estimates the likelihood of each of the regions rather than that of the parameter itself. Our simulation results show that the proposed algorithms achieve good performance as compared with traditional ML estimators.