Fast kernel MUSIC algorithm

Estimating parameters of closely spaced multiple signals has received much attention in the fields of radar and sonar. The multiple signal classification (MUSIC) algorithm has been considered the most effective. However, MUSIC requires much computation because of the analysis of the MUSIC eigenspectrum (Ψ) and the eigenvalue analysis of the covariance matrix C to estimate the noise vectors. Shimotahira (see Proc. ICASSP., vol.2, p.909-12, 1995 and IEICE Trans. Fundamentals, vol.E79-A, 1996) proposed the kernel MUSIC algorithm (K-MUSIC) which shortens its signal processing time. However, it does not change the fundamental points; C is formed and all noise vectors are used for Ψ. The fast kernel MUSIC algorithm (F-K-MUSIC) is proposed. In F-K-MUSIC, all we have to do is to estimate a reduced number of noise vectors without the formation of C and to analyze Ψ with only one noise vector. The reduction of the processing time of F-K-MUSIC ranges from one hundredth to one thousandth of that of MUSIC were shown by numerical simulations