Pseudo-Bayesian Approach for Robust Mode Detection and Extraction Based on the STFT

Area of Science:

• Signal Processing
• Data Analysis
• Mathematical Modeling

Background:

• Nonstationary signals are complex mixtures of underlying components.
• Accurate separation of these components is crucial for signal analysis.
• Existing methods struggle with noise and precise component localization.

Purpose of the Study:

• To develop a robust algorithm for disentangling multicomponent nonstationary signals.
• To accurately estimate the instantaneous frequency of individual signal modes.
• To enhance signal reconstruction quality in the presence of additive noise.

Main Methods:

• A novel pseudo-Bayesian algorithm for instantaneous frequency estimation.
• A detection algorithm to define component time-frequency regions.
• A new reconstruction approach using nonbinary band-pass synthesis filters.

Main Results:

• Successful disentanglement of nonoverlapping multicomponent signals.
• Accurate estimation of instantaneous frequencies for signal modes.
• Enhanced signal reconstruction quality compared to state-of-the-art methods.

Conclusions:

• The proposed pseudo-Bayesian approach offers effective signal disentanglement.
• The method demonstrates robustness against additive noise.
• Validated through experiments with synthetic and real-world data.