An exponential approach to signal parameter estimation
Emmanuel Racine, Dominic Grenier · 2012
This paper presents a general parameter estimation theory applicable in scenarios for which an observable signal expresses itself as a sum of independent random processes. The principle consists of evaluating the expected value of an exponential function of the observable signal, and finding the function parameter values for which the resulting expression vanishes. The proposed approach requires knowledge of the statistical distribution of the signals of interest, and may not apply to every type of signals. However, it proves immune to symmetric Gaussian noise (in the case of complex signals) and has the potential to identify more sources than sensors with no theoretical limit. An application example is provided as a means of evidencing the advantages of the theory.