Improved MUSIC using uniform subarrays

Performance analysis shows the asymptotic optimality of the MUSIC technique applied to bearing estimation problems for a sufficiently large number of sensors and not fully-coherent sources. This implies that a large number of covariance lags has to be computed; moreover the computational load of the eigendecomposition of large covariance matrices may be too severe for practical applications. With reference to uniformly spaced linear arrays (ULAs) we show that the accuracy gain associated to an increased number of sensors can be alternatively obtained by applying the MUSIC technique to particular configurations of pairs of ULAs, referenced to as subarrays, using a significantly smaller number of sensors. It is also shown that the accuracy loss of the proposed method, w.r.t. a full ULA covering the same array aperture, can be minimized by varying the distance between the two subarrays. The provided simulation results shows the applicability of the proposed method