A multi-domain graph-integrated neural framework for robust acoustic anomaly detection under adverse environmental conditions
Ayan Sar, Sumit Aich, Pranav Singh Puri, Sampurna Roy, Tanupriya Choudhury, Lubna Abdelkareim Gabralla, Minakshi, Bhupesh Kumar Dewangan · 2026
Read A multi-domain graph-integrated neural framework for robust acoustic anomaly detection under adverse environmental conditions by Ayan Sar, Sumit Aich, Pranav Singh Puri, Sampurna Roy, Tanupriya Choudhury, Lubna Abdelkareim Gabralla, Minakshi, Bhupesh Kumar Dewangan (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13433785.txt.
Description
Ayan Sar and colleagues build a neural framework for spotting unusual sounds in noisy, real-world settings, for security, defence, and city safety. Current sound models falter across noise, rare events, and changing distances, limiting their use outside the lab.
Their multi-domain, graph-integrated design unifies sound, time, and phase views through three encoders, wavelet, gammatone, and complex spectrogram, fused by a hierarchical transformer. Graph features grounded in sound-propagation physics add robustness, while a memory-boosted contrastive module sharpens rare-event recognition. Multi-task training covers classification and related goals, giving safety systems a sturdier ear for anomalies in harsh soundscapes.
Catalog record
| CREATOR | Ayan Sar, Sumit Aich, Pranav Singh Puri, Sampurna Roy, Tanupriya Choudhury, Lubna Abdelkareim Gabralla, Minakshi, Bhupesh Kumar Dewangan |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 270 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |