A recursive filter-based algorithm for maximum likelihood localisation of narrow-band autoregressive sources

Many conventional methods of source localization rely upon the offline maximization of a full data loglikelihood function. These functions are often complicated and difficult to maximize. Further, the computational burden in source localization via this form of optimization will typically depend upon the number of sensors in the sensor array, the number of signals whose directions are being estimated and the length of the measurement data set. Morever, standard schemes such as the EM algorithm, (see [9]), are not recursive. In this article we apply a recent recursive maximum likelihood estimation scheme, [3], to compute an estimate of the steering matrix for a passive uniform linear sensor A computer simulation is provided and performance is compared to the classical full data log likelihood function method.