Simplex based MUSIC for solving coherent signal

Abstract

An optimized MUSIC (MUtilple SIgnal Classification) is proposed, basing on Simplex search method. It can largely improve the scan speed of the conventional MUSIC and avoid the singularity of the Hessian matrix of the gradient based optimization methods. It also provides the possibility to distinguish the coherent signal with the acceptable time cost, in this case, we proposed an enumeration based coherence solver. Both simulations and experiments are carried out for RF localization purposes. The results shows that the proposed methods has faster computational speed with same numerical precision as the conventional MUSIC method does, which also confirms the feasibility of the proposed coherence solver.