Broadband source localization by regularization techniques
Bouchra Senadji, Yves Grenier · IEEE International Conference on Acoustics Speech and Signal Processing · 1993
The authors propose a new method for broadband source localization when few data are available. They aim to perform localization at a single frequency f/sub 0/ and to cope with the small amount of information drawn from the data. The first step is a frequency-dependent autoregressive moving average (ARMA) modeling of the signals coming from an array of sensors. The idea is to exploit the frequency variation of the ARMA vectors and consider it as a priori information on the vectors. Regularization techniques are then applied to optimize two criteria respectively representing the data and the a priori information. A regularization parameter is used to quantify the relative importance of the two criteria. The optimization of the regularized criterion leads to the estimation of the ARMA vector at frequency f/sub 0/ by solving a bloc tridiagonal system.>