Invariant Tests For Spatial Stationarity Using Covariance Structure

Abstract

We propose new tests for validating the ULA model based on the received array data. These tests use the structure of the covariance matrix of the data to decide whether the received data is spatially stationary. Invariance principles are used to ensure that the test has constant significance (equivalent to a CFAR test) when the model is indeed true by eliminating the dependence on the covariance. In addition, a modification of the test is proposed for verifying stationarity within a specified subspace, and heuristics are presented for distinguishing propagation effects from deviation from the ULA model.