Distributed spectral estimation using fuzzy set theory

A novel method to combine frequency estimates in a distributed processing scheme using fuzzy set theory is developed. The observed signal is assumed to be a sum of sinusoidal signals corrupted by an additive random noise process where the distribution of the noise process is assumed to be a contaminated Gaussian distribution, which is a more practical model than a pure Gaussian assumption. Distributed multisensor processing, or decentralized processing, in which each sensor estimates the frequencies of the observed signal independently and sends the results to the central process, is used. At each sensor, a robust frequency estimation method is used. At the central process, or the fusion center, frequency estimates are combined using fuzzy set theory. It is found that this method performs much better than conventional methods. Numerical simulations are given to confirm the performance of this method