[BCH17a] Higher order direction finding from rectangular cumulant matrices: the rectangular 2q-MUSIC algorithms

Revue Internationale avec comité de lecture : Journal Signal Processing, vol. 133, pp. 240-249, 2017, (doi:j.sigpro.2016.10.020.)

Mots clés: Cumulants, Virtual array, Direction finding, 2q-MUSIC, Rectangular arrangements, Non-Redundant

Résumé: Fourth-order (FO) high resolution direction finding methods such as 4-MUSIC have been developed for more than two decades for non-Gaussian sources mainly to overcome the limitations of second order (SO) high resolution methods such as MUSIC. In order to increase the performance of 4-MUSIC in the context of multiple sources, the MUSIC method has recently been extended to an arbitrary even order 2q (q ≥ 2), for square arrangements of the 2qth-order data statistics, giving rise to the 2q-MUSIC algorithm. To further improve the performance of 2q-MUSIC, the purpose of this paper is to extend the latter to rectangular arrangements of the data statistics, giving rise to rectangular 2q-MUSIC algorithms. Two kinds of rectangular arrangements, corresponding to redundant and non-redundant arrangements are considered. In particular, it is shown that rectangular arrangements of the higher order (HO) data statistics achieve a trade-off between performance and maximal number of sources to be estimated. These rectangular arrangements also lead to a complexity reduction for a given level of performance, which is still increased by non-redundant arrangements of the statistics. These results, completely new, open new perspectives in HO array processing.


@article {
title="{Higher order direction finding from rectangular cumulant matrices: the rectangular 2q-MUSIC algorithms}",
author="H. Becker and P. Chevalier and M. HAARDT",
journal="Signal Processing",