Coprime sampling and the music algorithm

A new approach to super resolution line spectrum estimation in both temporal and spatial domain using a coprime pair of samplers is proposed. Two uniform samplers with sample spacings MT and NT are used where M and N are coprime and T has the dimension of space or time. By considering the difference set of this pair of sample spacings (which arise naturally in computation of second order moments), sample locations which are O(MN) consecutive multiples of T can be generated using only O(M + N) physical samples. In order to efficiently use these O(MN) virtual samples for super resolution spectral estimation, a novel algorithm based on the idea of spatial smoothing is proposed, which can be used for estimating frequencies of sinusoids buried in noise as well as for estimating Directions-of-Arrival (DOA) of impinging signals on a sensor array. This technique allows us to construct a suitable positive semidefinite matrix on which subspace based algorithms like MUSIC can be applied to detect O(MN) spectral lines using only O(M + N) physical samples.