I. Introduction
The direction of arrival (DOA) estimation can be extensively applied in the field of radar, sonar, communication and navigation, and it is one of the most important and active subjects of array signal processing. By now there have been many DOA estimation methods in the literature, which can be classified into two kinds, beamforming techniques and subspace-based methods [1]–[6]. The Capon's beamformer [3], also known as the minimum variance distortionless response (MVDR) filter, can alleviate the limitations of the conventional beamformer on resolving two signals spaced closer than a beamwidth. Also, the multiple signal classification (MUSIC) algorithm has been regarded as one of the best bearing estimation methods since its publication by Schmidt in 1979 [4]. It is known that MUSIC can obtain an arbitrary high accuracy as well as high resolution for equal power sources if the number of snapshots and the SNR of sources are sufficiently high. Root-MUSIC [5] which is used in uniform linear array only can obtain better performance than MUSIC. Nevertheless, when the powers of different sources have obvious differences, the resolution and accuracy of multiple sources will be deteriorated accordingly [1]–[2], [6]. Normally, the eigenvalues in eigenvalue spectrum and the peaks in the spatial spectrum will become smaller and difficult to identify for the low power sources due to the mutual effect among different sources.