Performance weighted blended spectral estimation on experimental seaglider data

A common problem in underwater acoustics is the non-parametric estimation of a power spectrum from time series data with windowed Fourier transforms. If characteristics of the environment such as signal to noise ratio or the frequency of loud line components are known then an appropriate window can...

Full description

Saved in:
Bibliographic Details
Published in:The Journal of the Acoustical Society of America 2020-10, Vol.148 (4), p.2545-2545
Main Authors: Tucker, Jeff, Wage, Kathleen E., Van Uffelen, Lora
Format: Article
Language:English
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A common problem in underwater acoustics is the non-parametric estimation of a power spectrum from time series data with windowed Fourier transforms. If characteristics of the environment such as signal to noise ratio or the frequency of loud line components are known then an appropriate window can be identified. When estimation is performed without a priori knowledge of the environment, an analyst often uses an ensemble of different windowed Fourier transforms and synthesizes the results. When no analyst is available this process must be automated. The performance weighted blended~(PWB) spectrum estimator automates the work of an analyst by weighting each estimator in an ensemble based on its performance and summing them to create a new estimator that performs as well or better than each estimator in the ensemble regardless of the characteristics of the environment [Tucker et al., IEEE UASP Workshop (2019)]. This talk summarizes the PWB estimator, and compares it to existing estimators when applied to passive acoustic data collected from a hydrophone mounted on a Seaglider autonomous underwater vehicle. Additionally, the talk presents modifications to the original PWB estimator that offer improved robustness in the presence of loud transient signals, e.g., noise from glider hardware. [Work supported by ONR.]
ISSN:0001-4966
1520-8524
DOI:10.1121/1.5147066