This paper describes a method for locating multiple wideband sources using the signals received by an array of sensors. In the present approach, the problem of source location is treated as a frequency dependent estimation problem. Each wideband source is received by the array as a vector signal. The spatial spectrum of this signal has a linear frequency modulation (with respect to the usual frequency) which is characteristic of the source location. The frequency modulation may be recovered by frequency dependent modeling over the entire frequency band. The models are estimated by a technique of "model expansion," which consists in the replacement of the data by their projections on a basis of functions of frequency. An invariant expanded model is fitted to these projections. It expresses the evolution of a spatial model with respect to frequency. An estimator based upon an eigen-decomposition of a covariance matrix built from the observations, is discussed. This estimator leads to wideband versions of the high resolution estimators (minimum norm solution and MUSIC). Simulations will conclude the paper and compare this estimator with the coherent signal subspace method.
展开▼
