Peak Extraction and Partial tracking of Music signals using Kalman filtering

In this paper we propose a partial tracking method for music signals based on Kalman filtering. We first introduce a novel technique for detection of peaks in spectral representations of music signals. We also introduce different evolution models for our Kalman filter based on the shape of frequency and power partials in different classes of melodic instruments. Parameters of these models are estimated using a large database of music signals. We analyze the performance of our tracker through a comparison with another method and also by observing its effectiveness in the presence of crossing partials and vibrato. The problem of missing peaks and the contribution of a backward tracker are also discussed.