PNNL

Abstract

Machine learning classification and regression algorithms were applied to calibrate the localization errors of a time-difference-of-arrival (TDOA) based acoustic sensor array used for tracking salmon passage through a hydroelectric dam on the Snake River, Washingon, USA. The locations of stationary and mobile acoustic tags were first tracked using the approximate maximum likelihood (AML) algorithm. Next, ensembles of classification trees successfully identified and filtered data points with large localization errors. This prefiltering step allowed the creation of a machine-learned regression model function, which decreased median distance error by 50% (34%) for the stationary (mobile) tracks, and extended the previous range of submeter localization accuracy from 100 meters to 250 meters horizontal distance from the dam face. Errors in the depth direction were greatly decreased, falling by 82% and 92%, respectively, for the stationary and mobile tracks. These same methods would have application to the calibration of error in any TDOA-based sensor network with a steady environment and array configuration.