A Hands-On Tutorial on the Underwater Acoustic Channel Library

About This Tutorial

Reproducible evaluation of underwater acoustic receivers requires the ability to drive any candidate signal through a realistic channel and a realistic noise field, with full control over the experimental conditions. Most researchers solve this problem in private, with code and channel files that never leave the originating laboratory.

The Underwater Acoustic Channel Library, available at github.com/uwa-channels, removes this barrier. It provides eight measured channels, recorded in the North Atlantic, off Martha's Vineyard, across the eastern North Pacific, off Kauai, in a Norwegian fjord, off the coast of Japan, near Singapore, and in the Mariana Trench. Together they span transmission ranges from tens of meters to thousands of kilometers and center frequencies from 75 Hz to 25 kHz. The channel data is archived on Zenodo under doi:10.5281/zenodo.21287414.

The companion MATLAB and Python toolboxes implement three core operations: replay (pass any signal through a measured channel), noisegen (generate realistic ocean noise), and unpack (reconstruct the full time-varying impulse response from the compressed storage format).

This tutorial walks you through integrating the toolbox into your evaluation pipeline. You will leave with working code on your laptop that replays your own signals through several measured channels and adds realistic noise, and with a clear mental model of the underlying representations and assumptions.

Who Should Attend

Graduate students, postdoctoral researchers, early-career faculty, and practicing engineers working on underwater acoustic communications, signal processing, channel modeling, or machine learning for ocean signals. The tutorial is also suitable for newcomers from adjacent fields (terrestrial wireless, geophysics, machine learning) who want to access measured underwater data without first running their own sea trials.

Prerequisites

• Basic digital communications concepts (passband/baseband representations, multipath channels, equalization at a textbook level).
• Comfort with MATLAB or Python at the level of writing and debugging short scripts.
• No prior experience with the Library is assumed.

Learning Objectives

By the end of the tutorial, you will be able to:

• Navigate the Library's documentation site, download channels from Zenodo, and locate the appropriate toolbox and example scripts.
• Read and interpret the compressed channel file format, including the impulse-response tensor and the phase/delay-phase trajectory.
• Replay a passband signal of your own design through a measured channel and recover the received waveform at each receive element.
• Generate ocean noise consistent with measured statistics, including pink Gaussian spectra, spatial correlation across array elements, and impulsive heavy-tailed components.
• Unpack the compressed channel representation into a full time-varying impulse response and visualize its delay-Doppler structure.
• Locate and use the estimate repository to extract a channel from your own at-sea recordings.

Instructors

Schedule

Half-day single-track event (4 hours), held on the morning of Monday, September 21, 2026, within the OCEANS 2026 workshops and tutorials block. Please bring your laptop with the software pre-installed (see Setup Guide).

Time	Session
08:30 – 09:30	Introduction Overview of underwater acoustic channels: propagation physics, multipath and Doppler, time variability, and the channel estimation methodology underlying the Library.
09:30 – 10:10	Module 1: Replay Load a channel file, inspect its contents, generate a BPSK probe signal, call replay to drive it through the channel, and plot the received waveform, cross-correlation, and spectrum. Hands-on exercise with a channel and probe of your choosing.
10:10 – 10:30	Coffee Break ☕
10:30 – 10:50	Module 2: Ocean Noise Call noisegen in three configurations: pink Gaussian (17 dB/decade slope), spatially correlated Gaussian across a multi-element array, and impulsive symmetric alpha-stable. Each configuration is followed by a diagnostic exercise.
10:50 – 11:10	Module 3: Unpack & Visualize Use unpack to reconstruct the full time-varying impulse response, visualize its evolution as a delay-time waterfall, and identify physical structures (direct path, surface/bottom bounces, Doppler-induced delay drift).
11:10 – 11:40	Module 4: The Library at a Glance Overview of the eight hosted channels, their geographic and parametric coverage, and criteria for choosing a channel. Walkthrough of the documentation website, the Zenodo record, and the GitHub organization layout (matlab, python, estimate).
11:40 – 12:20	Module 5: The estimate Repository Walkthrough of the companion repository: estimator pipeline, probe signals and constraints, visualization scripts. Attendees with at-sea recordings can attempt a first estimation pass.
12:20 – 12:30	Wrap-Up Contribution workflow for submitting newly estimated channels, overview of the GitHub discussion forum, and Q&A.

Materials

All materials will be available here before and after the tutorial. Template scripts will be provided in both MATLAB and Python.

Pre-Tutorial Setup Guide

Please complete the following steps before arriving at the tutorial so that the in-room network is not a bottleneck. A detailed setup guide will be emailed to registered attendees two weeks before the event.

Option A: Python

1. Install Python 3.10 or later (python.org or via Anaconda/Miniconda).
2. Install the toolbox (includes all dependencies):

   pip install uwa-channels

3. Download the two announced channel files from Zenodo.
4. Verify your setup by running example_replay.py from the Python toolbox repository.

Option B: MATLAB

1. Install MATLAB (R2021a or later) with the Signal Processing Toolbox.
2. Install the toolbox by pasting this into the MATLAB Command Window:

   matlab.addons.install(websave([tempname '.mltbx'], 'https://github.com/uwa-channels/matlab/releases/latest/download/uwa-channels.mltbx'))

   This downloads the latest release and installs it as an add-on, so replay, noisegen, and unpack are on your path in every session. You can also download uwa-channels.mltbx from the latest release and double-click it.
3. Download the two announced channel files from Zenodo.
4. Verify your setup by running example_replay.

Option C: From a clone, and for Octave

Take this route if you want to read or modify the source, or if you are running Octave 9.0 or later, which cannot read .mltbx files.

1. Clone the repository:

   git clone https://github.com/uwa-channels/matlab.git
   cd matlab

2. Run install from that folder in MATLAB or Octave. This adds src and examples to the search path and saves the path for later sessions.
3. Under Octave, also install the signal and statistics packages.

Do not use Options B and C at once. If the add-on is installed and you also run install, two copies sit on the path and load order decides which one wins.

No local install? Run it in the browser

If you cannot install anything on your laptop, or your MATLAB license will not travel with you, both toolboxes still run.

Questions

Questions about the tutorial itself can go to Zhengnan Li. Questions about the toolboxes are best raised as an issue on GitHub, where the answer helps everyone.