Underwater acoustics

Synthetic passive and active sonar data for detection, classification and tracking models — shaped by the ocean the sensor actually operates in.

Request an unclassified sample

01 — The problem

Acoustic data is scarce, sensitive and tied to the day it was recorded.

Measured recordings are expensive to collect, often classified, and specific to the water, season and platform on the day. Models trained on them can struggle when any of those conditions change.

Physics-based synthetic data lets you vary the ocean deliberately — sound-speed profile, bathymetry, sea state, shipping density — and see how your model holds up.

02 — The physics we model

From source to sensor.

Propagation

Sound-speed profiles, bathymetry, and seabed and surface interaction, using established ocean-acoustic models.

Ambient noise

Shipping, wind, rain and biological noise, varied by location and season.

Source signatures

Tonal and broadband components that change with speed and operating state.

Platform and array

Array geometry, beamforming, self-noise and the receiving platform's motion.

Active sonar

Target strength, reverberation and multipath for echo-ranging systems.

Environmental variation

Systematic sweeps across conditions, so coverage gaps are visible rather than assumed.

03 — Outputs and labels

Data in the form your pipeline expects.

Passive: Spectrograms, modulation spectra and bearing–time records.
Active: Echo returns and range–bearing maps.
Raw: Element-level time series for your own processing chain.
Labels: Contact class, bearing, range, signal-to-noise ratio and full environmental metadata.
Formats: HDF5 or WAV with JSON metadata [confirm]

Illustrative only — not measured or modelled data.

04 — How we validate

Evidence, not assertion.

01

Compare

Synthetic outputs are compared with measured recordings from comparable conditions, where available.
02

Quantify

The synthetic-to-real gap is reported with the metrics behind it, including train-on-synthetic, test-on-real performance.
03

Document

Validated ranges and known limits are recorded in the data card that ships with every dataset.

Read about our assurance approach →

05 — Applications

Contact detection and classification Maritime domain awareness Uncrewed underwater vehicle autonomy Seabed object detection Test and evaluation of acoustic processing

A note on classification. Public samples are generic and unclassified. Platform-specific signatures are produced only within appropriate security arrangements.