Performance analysis of ESPRIT algorithm on a single broadband signal

In this paper, the problem of bearing estimation of a single wideband source is addressed. ESPRIT is a subspace based method which was originally developed by Roy and Kailath for direction of arrival estimation of narrow-band emitter signals. To use this method for wide-band signal, the wide-band signal is first decomposed into several narrow band frequency bins using Fourier Transform, and ESPRIT is subsequently applied separately on each frequency bin. This is a computationally costly method but shows promising results. Simulations are used to describe the performance of the said algorithm, when the number of frequency bins is changed. It is also shown that the variance of estimates is increased as the angle of arrival from the broad-side of a uniform linear array is increased.