A generalized AR operator for non-uniform noise compensation in DOA signal enhancement

Abstract

In some DOA applications, the M sensors of the array can be affected by spatially uncorrelated white noise having power levels different from sensor to sensor. Such a noise power unbalance induces severe bias errors of DOA estimates performed by eigenspace based methods. In some instances, an estimate of the ratio between the noise variances of the sensors is available, so that a noise pre-compensation procedure can be conceived. For this purpose, a generalized AR estimator is considered in connection with the signal enhancement approach. Simulation examples illustrate the method.<>