Estimating a Function from Noisy Sensor Data: A Factor Graph Approach

Abstract

The combination of graphical models and belief propagation algorithms has found wide acceptance in the design of communication systems. We extend the general framework of joint source-channel decoding on graphs to account for estimation problems in which the goal is not to decode the entire data set but to estimate a function of the transmitted data. This problem is deemed relevant e.g. in the context of wireless sensor networks.