Transform-based covariance differencing approaches to the estimation of multiple-signals in a multi-sensor environment

Eigendecomposition-based methods developed to date require that the additive noise be spatially white, i.e. of equal power and uncorrelated from sensor to sensor. A set of transform-based covariance differencing approaches is introduced that eliminates the effects of unequal noise power on eigenvalues and, at the same time, increase the resolution of the procedure in comparison with the standard methods. The performance of this approach under different conditions is examined through simulations, and the results are compared to the results obtained by using the standard MUSIC algorithm.<
>