Robust Kalman Filtering with Application to Tracking of Partials in Music Signals

In this paper we propose a novel method for tracking of partials in music signals based on a robust Kalman filter. This tracker is based on a regularized least-squares approach that is designed to minimize the worst-possible regularized residual norm over the class of admissible uncertainties at each iteration. This method promises improved tracking capabilities, compared with the conventional Kalman filter, which was proposed before. The model parameters that have been estimated for different frequencies are now considered as bounded uncertainties. Unlike the conventional Kalman tracker, the performance of this tracker is not influenced by the magnified track variations in higher frequencies.