Wideband source localization by space-time MUSIC subspace estimation

Abstract

Accurate estimation of the direction of arrivals (DOAs) of multiple wideband signal sources by sensor arrays is of paramount importance in recent developments of Ultra-Wide Band (UWB) and MIMO communication systems, acoustic applications, ultrasound, beside classical radar and sonar sensing. The array model changes with frequency. Narrowband analysis is not suited for short duration and, more in general, non-stationary sources. Most existing wideband direction finding algorithms are based on sensor output channelization (frequency binning) and neglect correlations among frequency bins, intra-bin finite bandwidth effects and spectral leakage that may create ghost sources during signal subspace estimation and impair the consistency of DOA estimators at high signal to noise (SNR) ratios. In this paper, a minimum leakage MUSIC-based estimator of subband signal subspaces from the space-time array covariance is introduced. Resulting subspace estimates can be fed to any frequency domain Maximum Likelihood (ML), Weighted Subspace Fitting (WSF) or focusing algorithm for final DOA estimation. Realistic simulations demonstrate the superior performance of the new estimator in difficult environments.