Charrelation-assisted covariance fitting
Arie Yeredor, Alon Slapak · 2012
Covariance fitting is a commonly used approach in array processing for estimating the power of signals impinging on a sensors array, and/or for refining estimates of the array's steering vectors. In this work we consider the possibility to further refine these estimates using a recently proposed generic statistic - called the Charrelation matrix, similar in form and in structure to the covariance matrix, but generally carrying information beyond second-order. The charrelation matrix and the statistics of its sample-estimate depend on the selection of a parameters-vector called "processing-point". As we show in here, the use of charrelation matrices taken at one or more processing-points as a substitute to the covariance (which is the charrelation matrix taken at an all-zeros processing-point), can yield significant improvement in the resulting estimates of the steering-vectors.