Instantaneous Frequency and Amplitude Estimation in Multicomponent Signals Using an EM-Based Algorithm

LEGROS, QUENTIN; FOURER, DOMINIQUE; MEIGNEN, SYLVAIN; COLOMINAS, MARCELO ALEJANDRO

IEEE Transactions on Signal Processing

Institute of Electrical and Electronics Engineers Inc.

Año: 2024 vol. 72 p. 1130 - 1140

This paper addresses the problem of estimating theinstantaneous frequency (IF) and amplitude of the modes composinga non-stationary multi-component signal in the presenceof noise. A novel observation model for the signal spectrogramis developed within a Bayesian framework to handle intricateconfigurations involving noise or overlapping components. Themodel parameters are estimated using a stochastic variant ofthe Expectation-Maximization algorithm, bypassing the computationallyexpensive joint parameter estimation from the posteriordistribution. We then design an algorithm for instantaneousamplitude and frequency estimation that accounts for overlapand amplitude variations of the components. To assess theperformance of the proposed method, we conduct experimentson both real-world and simulated signals, involving separated orcrossing modes. The benefits of our method in terms of efficiencycompared with several state-of-the art techniques appear to besignificant in that latter case, but also when the amplitude of thecomponents are varying across time.