Chapter 11Advanced Processing for DOA Estimation1
11.1.1. Standard processing for DOA estimation
Conventional direction of arrival (DOA) estimation techniques, using either amplitude (Watson-Watt), phase (Doppler, interferometry [MUL 87]) or both amplitude and phase of observations (beamforming, ecartometry), were developed before the 1960s. They were mainly built to process situations characterized by a unique source with a main path and strongly degrade in the presence of multiple sources and/or multiple propagation paths of a given source. The scarce conventional methods, such as the beamforming method, able to accommodate multiple sources or multiple propagation paths generally generate, in these cases, a bias in the DOA estimation and are limited in resolution by the 3dB beamwidth of the array, called the Rayleigh distance.
To overcome the previous limitations, super-resolution DOA estimation techniques, such as Capon’s [CAP 69] or Auto-Regressive methods [BUR 72], were developed from the end of the 1960s. These techniques come from spectral analysis and are related to the concept of generalized matched filter. These techniques allow us to separate sources and/or propagation multi-paths inside the Rayleigh distance provided their number P does not exceed N – 1 (AR) or N (Capon), where N is the number of antennas. Nevertheless, these techniques generally generate a bias in the DOA estimation in multiple sources context, especially for weak sources. ...