Underwater passive acoustic localization of Pacific walruses in the northeastern Chukchi Sea
| Autor: | Brendan P. Rideout, Stan E. Dosso, David E. Hannay |
|---|---|
| Rok vydání: | 2013 |
| Předmět: |
Sound Spectrography
Time Factors Acoustics and Ultrasonics Meteorology Oceans and Seas Transducers Monte Carlo method Marine Biology Arts and Humanities (miscellaneous) Bayesian information criterion Maximum a posteriori estimation Animals Computer Simulation Underwater Swimming Uncertainty analysis Hydrophone Principle of maximum entropy Uncertainty Bayes Theorem Signal Processing Computer-Assisted Acoustics Geodesy Linear Models Walruses Vocalization Animal Akaike information criterion Monte Carlo Method Alaska Algorithms Geology Environmental Monitoring |
| Zdroj: | The Journal of the Acoustical Society of America. 134:2534-2545 |
| ISSN: | 0001-4966 |
| DOI: | 10.1121/1.4816580 |
| Popis: | This paper develops and applies a linearized Bayesian localization algorithm based on acoustic arrival times of marine mammal vocalizations at spatially-separated receivers which provides three-dimensional (3D) location estimates with rigorous uncertainty analysis. To properly account for uncertainty in receiver parameters (3D hydrophone locations and synchronization times) and environmental parameters (water depth and sound-speed correction), these quantities are treated as unknowns constrained by prior estimates and prior uncertainties. Unknown scaling factors on both the prior and arrival-time uncertainties are estimated by minimizing Akaike's Bayesian information criterion (a maximum entropy condition). Maximum a posteriori estimates for sound source locations and times, receiver parameters, and environmental parameters are calculated simultaneously using measurements of arrival times for direct and interface-reflected acoustic paths. Posterior uncertainties for all unknowns incorporate both arrival time and prior uncertainties. Monte Carlo simulation results demonstrate that, for the cases considered here, linearization errors are small and the lack of an accurate sound-speed profile does not cause significant biases in the estimated locations. A sequence of Pacific walrus vocalizations, recorded in the Chukchi Sea northwest of Alaska, is localized using this technique, yielding a track estimate and uncertainties with an estimated speed comparable to normal walrus swim speeds. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|