Blind channel estimation of time-varying underwater acoustic waveguide impulse responses

Abstract

Some techniques for underwater passive acoustic localization make use of estimates for the direct and/or interface-reflected acoustic arrival times of a source at one or more underwater hydrophones. This estimation task can be difficult for non-impulsive sources (e.g., humpback whales) due to early arrivals masking later ones. In linear system analysis, the impulse response (IR) of a system is the system output when the input is an impulse (i.e., a short duration, large bandwidth signal), and expresses how a source signal interacts with the environment to yield the output waveform (e.g., that recorded by a hydrophone). Recordings of a relatively impulsive vocalization (e.g., a walrus knock), given that they approximate the IR between the walrus and hydrophone, may facilitate arrival time estimation. IR estimation for unknown, non-impulsive calls is more challenging, particularly for time-varying channels. Blind channel estimation is the process of estimating the set of IRs between a single (unknown) source and multiple receivers, and can potentially help estimate direct and interface-reflected arrival times for non-impulsive marine mammal vocalizations since the IRs can be analyzed rather than waveforms/spectra. In this paper, we use simulations to explore the importance of ocean surface gravity waves on blind estimation of acoustic IRs.