analiza aiqbench
This research advances bioacoustic monitoring by shifting from fixed-window analysis to time-frequency localized classification of individual vocalizations using foundation models. The proposed Dual-Context Classifier leverages both local events and global acoustic context to achieve high accuracy while reducing reliance on labeled data through self-supervised pre-training. This approach enables more precise ecological studies and long-term soundscapes monitoring without diluting predictions across multiple overlapping calls.
Localized time-frequency representation learning for bioacoustic classification in complex soundscapes
arXiv:2502.13440v2 Announce Type: replace-cross Abstract: Prevailing bioacoustic classifiers assign species labels to fixed time-frequency windows rather than to individual vocalizations. When multiple vocalizations occur within the same window, predictions cannot be unambiguously linked to specific calls, which limits analyses at the level of individual vocalizations. This work introduces a framework for time-frequency localized bird classification. A Local-Context Classifier (LCC) identifies species from localized time-frequency events (TFEs), while a Dual-Context Classifier (DCC) combines local and global acoustic context through a fine-tuned bioacoustic foundation model. On an in-distribution dataset from Singapore comprising 306 vocalization classes from 103 bird species, the LCC achieves an F1-micro score of 79.3%, while combining local and global context through the DCC yields the highest overall performance (94.6%). To reduce labeled data requirements, the LCC is pre-trained via self-supervised contrastive learning, achieving an 18.8% relative gain on an out-of-distribution dataset. A focused evaluation on continuous soundscape recordings further demonstrates the potential of the framework for long-term monitoring applications. By preserving the time-frequency localization of individual vocalizations, the proposed framework supports both ecological monitoring and vocalization-level studies of animal acoustic behavior.
Abstract:Prevailing bioacoustic classifiers assign species labels to fixed time-frequency windows rather than to individual vocalizations. When multiple vocalizations occur within the same window, predictions cannot be unambiguously linked to specific calls, which limits analyses at the level of individual vocalizations. This work introduces a framework for time-frequency localized bird classification. A Local-Context Classifier (LCC) identifies species from localized time-frequency events (TFEs), while a Dual-Context Classifier (DCC) combines local and global acoustic context through a fine-tuned bioacoustic foundation model. On an in-distribution dataset from Singapore comprising 306 vocalization classes from 103 bird species, the LCC achieves an F1-micro score of 79.3%, while combining local and global context through the DCC yields the highest overall performance (94.6%). To reduce labeled data requirements, the LCC is pre-trained via self-supervised contrastive learning, achieving an 18.8% relative gain on an out-of-distribution dataset. A focused evaluation on continuous soundscape recordings further demonstrates the potential of the framework for long-term monitoring applications. By preserving the time-frequency localization of individual vocalizations, the proposed framework supports both ecological monitoring and vocalization-level studies of animal acoustic behavior.
Substantially revised, including new title and abstract. Main changes: expanded literature review; evaluation on an out-of-distribution dataset; additional BirdNET configurations; new dual-context classifier; ablations on TFE extraction and contrastive pre-training; simplified pipeline (autoencoder removed); restructured manuscript
Sound (cs.SD); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Audio and Speech Processing (eess.AS); Quantitative Methods (q-bio.QM)
arXiv:2502.13440 [cs.SD]
(or arXiv:2502.13440v2 [cs.SD] for this version)
Text integral sub licența deschisă a publicației · 240 cuvinte
Subiecte conexe
Predicting the Outcome of rTMS Depression Therapy using EEG Signals and CNN
arXiv · 28 iul. 2026
In-Context Time Series Classification with Random Convolutional Features
arXiv · 22 iul. 2026
Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
Latent Space · 8 oct. 2026
[AINews] The Future of Latent Space
Latent Space · 25 sept. 2026