Kalman filter for dynamic source power and steering vector estimation based on empirical covariances

Interferometric measurements correspond to sample covariance matrices of signals received by multiple sensors. In dynamic scenarios, such as radio astronomy imaging, the properties of these signals can vary over time, posing a significant challenge for study. This work addresses the issue of estimating the stochastic power and steering vector of signal sources from sample covariance measurements. A novel approach is proposed, introducing a non-standard Kalman filter designed to accommodate any noise and signal distribution, thereby broadening the Kalman filter’s applicability to situations with unknown measurement models. The effectiveness of this method is highlighted in the case of joint estimation of source power and direction of arrival through simulations using synthetic data.