Study On DOA Estimation Via Circular Arrays With Fourth Order Cumulants

The direction of arrival (DOA) estimation is one of the most important topics in array signal processing. Regarding the array statistics used in the DOA estimation, the fourth-order cumultants (FOCs) have been widely studied for their advantages including the (asymptotical) insensitiveness to Gaussian noise that can be used to suppress Gaussian noise without knowing or estimating the noise covariance and the ability of array aperture extension that can improve the DOA estimation resolution. On the other hand, the uniform circular arrays (UCAs) have a special array structure and superior direction finding performance. As compared with uniform linear arrays (ULAs), UCAs can estimate directions in the range of 360 degrees and have isodirectional estimation performance. It is meaningful to do the DOA estimation by using FOCs and UCAs, especially the coherent signal DOA estimation which is an important problem in communication and radar since multi-path propagation or spot jamming often occurs in a real environment.In this study, we use FOCs to do the DOA estimation based on the uniform circular arrays for independent signals and coherent signals in the unknown colored Gaussian noise, respectively. The main work includes:(1) The UCA array signal models are described for independent signals and coherent signals, respectively. Then, the linear transformation between the UCA array signal model and the ULA array signal model is studied. Further, we introduce the concept and the basic properties of the fourth order cumulants.(2) In the presence of colored Gaussian noise, several often-used FOC-based methods are studied for independent signals impinging on the UCAs, including MUSIC-like, MFOC-MUSIC, virtual-ESPRIT and RARE-Cumulant. The Gaussian noise suppression of these methods is examined through the computer simulations.(3) A new DOA estimation method is presented for coherent signals with a UCA. The presented method first applies the so-called mode-space transformation to represent the UCA data as the virtual ULA data which has the Vandermonde structure, then constructs special cumulant matrices, together with the spatial smoothing technique and the MUSIC-like subspace method, to obtain the DOA estimation of coherent signals. The simulation results validate the performance of the presented method.