Employing moments of multiple high orders for high-resolution underdetermined DOA estimation based on MUSIC

Several extensions of the MUltiple SIgnal Classification (MUSIC) algorithm exploiting high order statistics were proposed to estimate directions of arrival (DOAs) with high resolution in underdetermined conditions. However, these methods entail a trade-off between two performance goals, namely, robustness and resolution, in the choice of orders because use of high-ordered statistics increases not only the resolution but also the statistical bias. To overcome this problem, this paper proposes a new extension of MUSIC using a nonlinear high-dimensional map, which corresponds to the joint analysis of moments of multiple orders and helps to realize the both advantages of robustness and high resolution of low-ordered and high-ordered statistics. Experimental results show that the proposed method can estimate DOAs more accurately than the conventional MUSIC extensions exploiting moments of a single high order.