Novel eigenanalysis method for direction estimation

TL;DR: A new eigenanalysis-based technique for direction estimation (and for estimation of the parameters of superimposed exponential signals from multiexperiment noisy data) is introduced, which offers the performance of the maximum likelihood (ML) method at a modest computational effort, which is comparable to that associated with other eigen analysis-based techniques such as the MUSIC algorithm.

Abstract: A new eigenanalysis-based technique for direction estimation (and for estimation of the parameters of superimposed exponential signals from multiexperiment noisy data) is introduced. This novel technique, which is called MODE (method of direction estimation), offers the performance of the maximum likelihood (ML) method (the MODE and ML estimators coincide as the number of data samples increases) at a modest computational effort, which is comparable to that associated with other eigenanalysis-based techniques such as the MUSIC algorithm. Compared to the latter, MODE offers the advantage of better performance, especially in situations where the sources are highly correlated. The type of performance that can be achieved by MODE is illustrated by means of some numerical examples which also show, for comparison, the corresponding performance achieved by the MUSIC algorithm and a popular approximate ML algorithm.