Efficient computation of the music algorithm as applied to a low-angle elevation estimation problem in a severe multipath environment

Subarray spatial smoothing of the covariance matrix decoheres a signal from its multipath. Since computations grow as fast as M³, where M is the subarray size, we should keep M to a minimum. Since the accuracy is found to improve as M increases and quickly asymptotes around a knee point, we suggest to choose M right after that point and use spatial smoothing to speed up MUSIC in general, even if signal coherence is of no concern. We also introduce a root finding algorithm working on a 2(M-1)th order polynomial as an efficient way of finding the peaks of the super resolution pattern.