Extension of ESPRIT method to unknown noise environments

With the assumption that noise correlation is spatially limited, it is proposed to use two subarrays to eliminate the effects of unknown noise (UN). To find the estimate of the signal subspace, canonical decomposition is used. Direction-of-arrival (DOA) estimation is then carried out by using the spatial invariance between the two subarrays, which is similar to the methodology of ESPRIT. The method provides a consistent estimator for all spatially band limited noise. Numerical simulations have verified that when the noise spectrum is moderately rough, UN-ESPRIT gives acceptable performance, while ESPRIT fails at even relatively high SNRs. On the other hand, for a given number of sensors, UN-ESPRIT has smaller aperture than ESPRIT