Sparse Array Construction using Marginal Fisher's Information

This paper presents a design algorithm for the effective construction of sparse sampling arrays. Using a form of Fisher's information, the algorithm can thin a sampling array in an optimal manner, such that the resulting array pattern has a minimal total integrated sidelobe level. The samples from the resulting array can thus be used to provide spectral estimates with minimized mean-squared error. A particularly powerful aspect of this algorithm is its relative speed, as compared to more traditional construction algorithms such as genetic algorithms. The resulting co-arrays from this method likewise exhibit very low redundancy.