Direction of arrival estimation based on weighted minimum mean square error

It is a challenging issue to design a robust broad-band direction-of-arrival (DOA) estimator at the cost of low computational complexity. Many algorithms accelerate localization while maintaining robustness by optimizing the grid search. Few algorithms pay attention to the cost function of DOA estimators. This paper presents a far-filed DOA estimator based upon a novel cost function for planar arrays. It has two distinctions from conventional ones. On one hand, an explicit solution of DOA is represented as a linear function of all time-difference-of-arrivals (TDOAs), which leads to a considerable reduction of computational complexity. On the other hand, the reliability of each TDOA is considered, which relieves the outlier influence. The proposed algorithm was evaluated in a simulated environment. Its robustness is close to SRP-PHAT, but significantly outperforms GCC-PHAT. It runs about 5 times faster than GCC-PHAT, and 30 times faster than SRP-PHAT on general desktop computers for a 9-element array.