DOA estimation of two targets using beamformer based methods with application to automotive radar
Javier Cordero Hernandez · 2011
Direction-of-arrival (DOA) estimation of two targets plays an important role in automotive radar. Two cases are distinguished: when the targets are closely spaced and the conventional beamformer is not able to resolve them, and when the targets are widely spaced and the beamformer is able to resolve them. In the first case, accurate estimates can be obtained using high-resolution techniques. In the second case, estimates are typically biased. Automotive radar applications demand real-time processing and therefore the computational cost has to be addressed. For the resolved scenario, we propose a procedure to reduce the bias of the beamformer estimates, thus avoiding the use of iterative algorithms. The final estimates are obtained after applying a correction term, which is calculated off-line and stored in a look-up table. For the non-resolved scenario, we consider a practicable implementation of the maximum likelihood estimator. A simplified version of the cost function is used to reduce the complexity. The peak location from the beamformer can also be used to delimit the search range. The results of the mentioned methods are compared with other iterative algorithms, in terms of performance and computational cost. Applying the correction factors, the bias of the beamformer estimates are successfully reduced, making them accurate enough for the automotive radar application. The simplified implementation of the ML cost function reduces significantly the computational cost, allowing its use in real-time applications. Moreover, the performance obtained is also within the acceptable range for the automotive radar application, even for narrow angular separations. A block diagram containing the proposed methods is finally given, which is proposed as a suitable DOA estimation system for the automotive radar application.