1. Introduction
Sensor arrays are prevalent in many applications such as radar, wireless communications, and sonar. By using multiple sensors, it is possible to separate multiple signals based upon their different direction of arrivals (DOAs). Beamforming and DOA estimation will allow for increased capacity in wireless communication ap-plication. Tracking applications require the ability to localize the source of the signals. The DOA estimation problem has been studied intensively and many good methods are available [1]. How-ever, most of these methods are designed to work exclusively for narrowband signals [2]. For signals whose bandwidth is significant relative to their center frequency, a few wideband DOA estimation techniques have been proposed (see [3] and the references therein). Most of the wideband methods decompose the signals into several narrowband frequency components through a series of filterbanks, estimate the spatial correlation matrix over each frequency component and use the structure of correlation matrices to arrive at a DOA estimate. There are two fundamental classes of methods to derive a DOA from the correlation matrices. First, the ‘incoherent’ methods process each frequency component independently and form a weighted average of DOA estimates over all frequency bins[4]. On the other hand, the ‘coherent’ methods form a global correlation matrix by averaging transformed versions of the individual correlation matrices for each frequency bin.