BROADBAND SOURCE LOCATION USING FREQUENCY-DEPENDENT MODELS.
Yves Grenier · 1988
SUMMARY. This paper describes a method for locating multiple broadband sources from the signals received on an array of sensors. In the present approach, the problem of source location is treated as a frequency dependent estimation problem. The signal emitted by each broadband source is received on the array as a vector signal, the spectrum of which has a linear frequency modulation (with respect to temporal frequency) in the spatial domain and an amplitude modulation in the temporal frequency domain. The frequency modulation is characteristic of the source location, and the amplitude modulation is a replicate of the source spectrum. The frequency modulation is recovered through frequency dependent AR (autoregressive) modeling on the entire frequency band. A least-squares estimator for this model is described, which extends the classical maximum entropy estimator used in the narrowband case. Another estimator based upon an eigendecomposition of some covariance matrix build over the observations, is also discussed. This estimator provides an extension to the narrowband Prony-Pisarenko estimator.