Maximum likelihood esimation of broadband noise source location for unknown signal and noise power spectral densities

摘要

Maximum likelihood estimator of broadband Gaussian noise source location in the presence of spatially uncorrelated background noise with adaptation to the unknown power spectral densities of noise and source signal is considered; the difference of the estimator from the analogous estimators derived earlier is discussed. The Cramer-Rao Lower Bound (CRLB) for characterization of location estimate variance is derived. The comparison of empirical variance estimates with CRLB and with empirical variance estimates in the case of conventional summation of narrow band array outputs is conducted. It is shown that the technique proposed provides better accuracy when the signal and noise power spectral densities are appreciable different. It is shown also that the effects of weak spatial noise correlation are negligible.