Time-Decoupled Spectral Modeling: Wide-Band Recalibration and Adaptive Fusion for Sound Event Detection
IEEE Signal Processing Letters · 2026 · 33 · 4206-4210
DOI: 10.1109/lsp.2026.3744304 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Signal Processing Letters, 2026. Verfasst von Qian Weng und 5 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Gefördert von: Industry-University-Research Cooperation Project of Fujian Province, Science and Technology Project for Police Development of Fujian Provincial Department, Fuzhou University and Yealink (Xiamen) Network Technology Co., Ltd..
Abstract (Originalsprache)
Convolutional recurrent neural networks (CRNNs) provide a compact solution for sound event detection (SED), but distant frequency bands interact only through successive local operations in their finite-depth convolutional front-ends. We propose Time-Decoupled Spectral Gating (TDSG), a parameter-efficient module that supplements CNN-based SED with wide-band spectral context. TDSG estimates a segment-conditioned spectral gate from a temporally aggregated channel–frequency descriptor while retaining the original frame-level feature for event localization. Element-wise adaptive fusion controls the strength of recalibration at each channel–time–frequency position. Controlled experiments on DESED and Urban-SED with CRNN and FDC-CRNN front-ends show higher mean detection scores over five runs per setting. On DESED, CRNN+TDSG improves the two Polyphonic Sound Detection Score scenarios (PSDS1/PSDS2) from 0.400/0.598 to 0.441/0.668 with 12.7% more parameters and 29.0% more FLOPs. On Urban-SED, it improves PSDS1/PSDS2 from 0.161/0.244 to 0.276/0.320. CRNN+TDSG also achieves higher mean PSDS than FDC-CRNN on both datasets with 54.9% fewer parameters and 60.4% fewer FLOPs. Ablation and timing analyses characterize the accuracy–complexity trade-off.
Förderung
- Industry-University-Research Cooperation Project of Fujian Province
- Science and Technology Project for Police Development of Fujian Provincial Department
- Fuzhou University and Yealink (Xiamen) Network Technology Co., Ltd.
Zitieren
APA
Weng, Q., Zeng, C., Lin, J., Kang, Y., Chen, M., & Jian, C. (2026). Time-Decoupled Spectral Modeling: Wide-Band Recalibration and Adaptive Fusion for Sound Event Detection. IEEE Signal Processing Letters, 33, 4206-4210. https://doi.org/10.1109/lsp.2026.3744304
Vancouver
Weng Q, Zeng C, Lin J, Kang Y, Chen M, Jian C. Time-Decoupled Spectral Modeling: Wide-Band Recalibration and Adaptive Fusion for Sound Event Detection. IEEE Signal Processing Letters. 2026;33:4206-4210. doi:10.1109/lsp.2026.3744304
BibTeX
@article{Weng2026,
author = {Weng, Qian and Zeng, Chenjie and Lin, Jiawen and Kang, Yuanxun and Chen, Mo and Jian, Cairen},
title = {Time-Decoupled Spectral Modeling: Wide-Band Recalibration and Adaptive Fusion for Sound Event Detection},
journal = {IEEE Signal Processing Letters},
year = {2026},
volume = {33},
pages = {4206--4210},
doi = {10.1109/lsp.2026.3744304}
}
EUVIMED