A new approach to the estimation of source motion parameters. II (Q1329482)

This paper deals with the estimation of source motion parameters in the general context of passive spatio-temporal processing. The temporal variations of the spatial-frequency rates \(\dot k_ t\) have been described by a Markovian model. Using this model, most of the practical situations (close or fast sources, etc.) can be considered. Following the general formalism of Part I of this paper [the authors, ibid. 33, No. 3, 287-314 (1993)], the estimation of the instantaneous sources spatial- frequency rates reverts to estimating a multiscale time-varying AR model whose poles are directly related to the \(\{\dot k_ i\}^ s_{i=1}\). For this purpose, Kalman Filtering constitutes a basic tool and allows us to consider a spatio-temporal source tracking. The hyperparameters are then estimated by maximizing a likelihood functional. This approach is then extended to multidimensional estimation. It is worth noting that this extension of spatio-temporal analysis may improve the performance of source tracking algorithms.