Direction of arrival estimation using weighted subspace fitting with unknown number of signal sources

In this paper, we propose a new approach on DOA estimation under the number of signal sources is unknown. This can be applicable even to the case of fully correlated sources for which existing algorithms such as AIC and MDL are not workable. AIC and MDL are well-known ldquoinformation theoretic criterionrdquo algorithms for estimating the number of sources observed by a direction finding array. In the proposed approach, the computational complexity for determining the source number is remarkably reduced by performing the estimation of DOAs and detection of the number of signals simultaneously. In addition, it can be applied to any geometry of array including ULA. The performance of the proposed approach is demonstrated using a numerical simulation. Results show that the probability of correct classification (PCC) of the number of incoming sources is satisfactory in most of simulations. Especially, the PCC of the approach for incoherent and coherent signals under the condition of two signals is very high so that it can classify signal sources reliably (i.e., accuracy >90%) at the SNR above 2 dB.