Enhancing Environmental Noise Management Through Siamese Convolutional Neural Network with Triplet Lost Function for Identification of Principal Sound Sources
Bibliographic record
Abstract
Identifying principal noise sources in distant fields is a challenging task due to the convergence of multiple sound sources and the degradation of acoustic signals over distance. This task becomes critical in industrial environments, where adherence to noise regulations is mandatory. Conforming to noise regulations can result in decreased productivity or even necessitate cessation of operations, significantly impacting the industry and the surrounding communities. Our goal, in response to these issues, is to develop a tool to facilitate a highly refined analysis of the causality relationship between industrially produced noise and operational productivity. Our approach is characterized by the robust identification of dominant noise sources in complex soundscapes, with a focus on specific sound events rather than locations or equipment types. To achieve this, we utilize Siamese Convolutional Neural Networks (SCNNs) with a triplet loss function. In our methodology, near-field captures from noisy equipment are used as anchors. The anchors are compared with both positive and negative instances captured by environmental noise monitoring stations in distant sound fields. Positive instances refer to those where the sound event produced by the equipment under study dominates, while negative instances are those where the anchor sound is present but not dominant. Preliminary results are promising, indicating that it is possible to discern among various near-field captures to identify the primary contributor to the noise detected outside an industrial plant. With real-time identification and loop back as the next goals, this study paves the way for more advanced noise monitoring, presenting its potential significant role in shaping noise management strategies within industrial contexts.
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 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 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) | 0.001 | 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 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".