The Esprit Algorithm With Higher-order Statistics

We address in this paper the bearing estimation problem of sources from array measurements for signal environments where the signal is non-Gaussian and the additive noise sources are colored (spatially correlated) Gaussian with unknown second-order statistics. The ESPRIT bearing estimation problem is reformulated using fourth- order cumulant matrices instead of autocorrelation matrices. By doing so, the fourth-order cumulant matrices of the additive colored Gaussian noises can be suppressed and therefore knowledge of the noise cross-correlation matrix becomes unnecessary. Simulation results are presented and performance comparisons are made between the fourth-order cumulantbased ESPRIT and iis equivalent second-order statistics-based version when the additive noise sources are colored Gaussian with unknown spatiid correlation matrix.