Prior-exploiting Direction-of-Arrival algorithms for partially uncorrelated source signals

• We investigate a recently proposed direction-of-arrival estimation algorithm.
• The statistical properties of the investigated estimator are derived, along with the relevant CRB.
• It is shown that the estimator is asymptotically statistically efficient given some benign assumptions.
• Through numerical Monte-Carlo simulations we additionally study the convergence behavior of the algorithm, and its robustness to model mismatch.

In this article, we investigate the performance of the recently proposed Direction-Of-Arrival (DOA) estimator POWDER (Prior Orthogonally Weighted Direction EstimatoR). The method is exploiting a specific form of prior information, namely that some DOAs are known, as well as that the correlation state between some of the source signals is known. In such scenarios, it is desirable to exploit the prior information already in the estimator design such that the knowledge can benefit the estimation of the DOAs of the unknown sources.Through an asymptotical statistical analysis, we find closed form expressions for the accuracy of the method. We also derive the relevant Cramér–Rao Bound, and we show the algorithm to be efficient under mild assumptions. The realizable performance in the finite sample-case is studied through numerical Monte-Carlo simulations, from which one can conclude that the theoretically predicted accuracies are attained for modest sample sizes and comparatively low SNR. This has the implication that the algorithm is significantly more accurate than other, state-of-art, methods, in a wide range of scenarios.