Direction-of-Arrival Estimation for Temporally Correlated Narrowband Signals

TL;DR

This paper investigates the impact of temporal correlation on direction-of-arrival estimation, showing bounds on performance and proposing an improved method that outperforms existing techniques in simulations.

Contribution

It demonstrates that the stochastic iid CRB bounds the temporally correlated CRB from above and introduces a modified IV-SSF method with enhanced accuracy.

Findings

Temporally correlated CRB is upper bounded by the iid CRB.

The proposed method reduces bias and error variance.

Simulation results confirm improved performance of the new method.

Abstract

signal direction-of-arrival estimation using an array of sensors has been the subject of intensive research and development during the last two decades. Efforts have been directed to both, better solutions for the general data model and to develop more realistic models. So far, many authors have assumed the data to be iid samples of a multivariate statistical model. Although this assumption reduces the complexity of the model, it may not be true in certain situations where signals show temporal correlation. Some results are available on the temporally correlated signal model in the literature. The temporally correlated stochastic Cramer-Rao bound (CRB) has been calculated and an instrumental variable-based method called IV-SSF is introduced. Also, it has been shown that temporally correlated CRB is lower bounded by the deterministic CRB. In this paper, we show that temporally correlated CRB is also upper bounded by the stochastic iid CRB. We investigate the effect of temporal correlation of the signals on the best achievable performance. We also show that the IV-SSF method is not efficient and based on an analysis of the CRB, propose a variation in the method which boosts its performance. Simulation results show the improved performance of the proposed method in terms of lower bias and error variance.