NW-MILO Acoustic Data Collection
Bibliographic record
Abstract
There is an enduring requirement to improve our ability to detect potential threats and discriminate these from the legitimate commercial and recreational activity ongoing in the nearshore/littoral portion of the maritime domain. The Northwest Maritime Information and Littoral Operations (NW-MILO) Program at PNNL’s Coastal Security Institute in Sequim, Washington is establishing a methodology to detect and classify these threats - in part through developing a better understanding of acoustic signatures in a near-shore environment. The purpose of the acoustic data collection described here is to investigate the acoustic signatures of small vessels. The data is being recorded continuously, 24 hours a day, along with radar track data and imagery. The recording began in August 2008, and to date the data contains tens of thousands of signals from small vessels recorded in a variety of environmental conditions. The quantity and variety of this data collection, with the supporting imagery and radar track data, makes it particularly useful for the development of robust acoustic signature models and advanced algorithms for signal classification and information extraction. The underwater acoustic sensing system is part of a multi-modal sensing system that is operating near the mouth of Sequim Bay. Sequim Bay opens onto the Straight of Juan de Fuca, which contains part of the border between the U.S. and Canada. Table 1 lists the specific components used for the NW-MILO system. The acoustic sensor is a hydrophone permanently deployed at a mean depth of about 3 meters. In addition to a hydrophone, the other sensors in the system are a marine radar, an electro-optical (EO) camera and an infra-red (IR) camera. The radar is integrated with a vessel tracking system (VTS) that provides position, speed and heading information. The data from all the sensors is recorded and saved to a central server. The data has been validated in terms of its usability for characterizing the signatures of small vessels. The sampling rate of 8 kHz and low pass filtering to 2 kHz results in an alias-free signal in the frequency band that is appropriate for small vessels. Calibration was performed using a Lubell underwater speaker so that the raw data signal levels can be converted to sound pressure. Background noise is present due to a nearby pump and as a result of tidal currents. More study is needed to fully characterize the noise, but it does not pose an obstacle to using the acoustic data for the purposes of vessel detection and signature analysis. The detection range for a small vessel was estimated using the calibrated voltage response of the system and a cylindrical spreading model for transmission loss. The sound pressure of a typical vessel with an outboard motor was found to be around 140 dB mPa, and could theoretically be detected from 10 km away. In practical terms, a small vessel could reliably be detected from 3 - 5 km away. The data is archived in netCDF files, a standard scientific file format that is "self describing". This means that each data file contains the metadata - timestamps, units, origin, etc. - needed to make the data meaningful and portable. Other file formats, such as XML, are also supported. A visualization tool has been developed to view the acoustic data in the form of spectrograms, along with the coincident radar track data and camera images.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.021 | 0.017 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".