Goal: Create fully automated classifiers for odontocete species that use information from whistles, clicks and the context of acoustic encounters and integrate these classifiers with PAMGuard and Ishmael software packages.
Time Period: May 2014 – September 2017
Our Role: to train and test fully automated classifiers for sounds produced by odontocete species in the northwest Atlantic, the temperate Pacific and the waters surrounding Hawaii.
The extensive use of Passive Acoustic Monitoring (PAM) as a tool for mitigation and monitoring of marine mammal populations has generated huge volumes of data, therefore automated methods for detecting and classifying sounds in the recordings are necessary. As such, Bio-Waves, Inc. is working to develop fully automated random forest classifiers for odontocete species. Typically, detectors and classifiers have been developed for one type of vocalization at a time (i.e., delphinid whistles or echolocation clicks), but using information from both whistles and clicks to classify odontocetes may significantly improve our ability to identify species acoustically. In addition, information about the context of acoustic encounters (location, acoustic behavior, etc.) may also provide useful inputs for classifiers. Bio-Waves will use information measured from whistles, clicks and the recording context to make classification decisions. Classifiers are being trained for odontocete species in the northwest Atlantic, the temperate Pacific and the waters surrounding Hawaii. If successful, the classifiers will be integrated with the PAMGuard and Ishmael acoustic data processing software platforms.
Collaborators/Partners: Southwest Fisheries Science Center, Pacific Islands Fisheries Science Center, Southeast Fisheries Science Center, Northeast Fisheries Science Center, Northwest Fisheries Science Center, Woods Hole Oceanographic Institution, Duke University, Southall Environmental Associates, Inc.
Sponsors: Office of Naval Research, U.S. Navy Living Marine Resources Program