Detection of Leopard seal (Hydrurga leptonyx) vocalizations using the Envelope-Spectrogram Technique (tEST) in combination with a Hidden Markov Model
Bibliographic record
Abstract
This paper describes a technique for the automated detection of leopard seal (Hydrurga leptonyx) vocalizations. Automatic detection of leopard seal calls within the Antarctic underwater soundscape is difficult because (a) the calls are frequently of low amplitude (b) the call duration is highly variable and (c) the frequency band overlaps with those of many other marine mammal vocalizations. However, humans easily distinguish leopard seal vocalizations from those of other marine mammals because of the calls' distinctive sound, which is a result of the pulsed structure of the leopard seal vocalizations. To exploit the unique temporal evolution of the pulse repetition rate (PRR) in high (HDT) and low (LDT) double trills, the Envelope-Spectrogram Technique (tEST) was developed. The extracted PRR feature allows detection of the target vocalizations even against a background of other marine mammal vocalizations. To handle the high variability of the calls' duration, the tEST algorithm was combined with a Hidden Markov Model (HMM) which is particularly well adapted to handle temporal variability. The developed HMM based detection algorithm worked rather reliably. The detection rate over a 4 day test period was high (72 %) although the signal to noise ratio (SNR) was low (< 10 dB). The number of false positive detections (12 %) was tolerable. Most of the false positives occurred during the period when R/V Polarstern was approaching the recording station and the SNR was temporarily < 0 dB. The detector worked 3 times faster than real-time and is therefore suitable for both off line biological research and time critical in-the-field applications, such as the detection of the presence of leopard seals in the context of human diver operations.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".