Wideband DOA estimation based on focusing signal subspace

The traditional wideband direction-of-arrival (DOA) estimation algorithms usually require the preliminary DOA estimations, which determine the accuracy of the final DOA estimations. To remedy the apparent performance degradation, this paper introduces a novel direction-of-arrival (DOA) estimation algorithm for wideband sources called the focusing signal subspace (FSS). This novel algorithm constructs a unitary focusing matrix which combines the eigenvector of signal subspace of the reference frequency and the eigenvector of signal subspace of other discrete frequencies with the constraint of Frobenius norm. The focusing matrix does not require the preliminary DOA estimations. Furthermore, the covariance matrix at each frequency is transformed to the focusing covariance matrix at the reference focusing frequency by using the focusing matrix. Experiments and simulation results show that focusing signal subspace outperforms the traditional rotational signal subspace and the test of orthogonality of projected subspace. FSS features a high resolution and a low root-mean-square error without the preliminary DOA estimations.