This repo processes data from the binaural arrays in the OrcaSound Network (consisting of 2 hydrophones in one node). As of August 2026, the binaural nodes are Orcasound Lab and Andrew's Bay.
This repo has one method of porting into the AWS buckets and pulling data. It's designed for data exploration, not huge data processing flows. Using orcasound_datapull_and_visualize.ipynb with orcasound_utils.py will allow the user to enter a time/date, hydrophone node, and desired directory for data download. The code will download the wav file for the defined time period (<1 hour to avoid crashing), and create a spectrogram accessibly via HTML link (<15min to avoid crashing).
Time Difference of Arrival (TDOA) localization is a technique used to determine the position of a sound source using multiple receivers. The method works by measuring the time differences between when a sound signal arrives at different hydrophone positions. Because each array only has 2 hydrophones, we can only get a source bearing from a single array. The hope is by using two nearby binaural arrays and crossing their bearings, we can get an estimated location.
-
DataVis: This folder should be used to pull data from the OrcaSound AWS database and process it to view specific time segments.
-
Annotation_Files: This folder has a lot of the manually determined metadata for different tests that were performed near this cable. If you want to play around with some of the sounds, these files will tell you where to look in the data. They are also used a lot in the code to automate TDOA processing for entire tests.
-
Localization: This folder works on localizing the calls based on their TDOAs. There are a few different iterations of localization somewhat split up as follows:
tdoa_AB.ipynb: source bearing estimate at Andrew's Bay only.tdoa_OSLab.ipynb: source bearing estimate at Orcasound Lab only.tdoa_cross_bearings.ipynb: tdoa at both nodes and then crossed to intersect at one location (hopefully!)