Particle filtering for dispersion curve estimation from spectrograms of acoustic signals

Abstract

In this work we develop a particle filtering method for dispersion curve extraction from spectrograms of acoustic signals, the goal being to obtain accurate representation of modal dispersion for geoacoustic inversion. The approach combines particle filtering with modeling of sound propagation in ocean environments to track dispersion curves of multimodal signals in noisy environments. ‘‘Crispy’’ estimates of dispersion curves are extracted that are not characterized by extensive uncertainty typically hindering geoacoustic inversion. Results are presented from both synthetic signals and real data collected in the East China Sea during the ASIAEX experiment. [Work supported by ONR.]