Research on Urban Sound Classification based on ConvNeXt-FECA Model
DOI:
https://doi.org/10.54097/bha9rp62Keywords:
Sound Classification, ConvNeXt, Frequency-enhanced Convolution Attention, Spectrogram, MFCCAbstract
With the increasing severity of urban sound pollution, efficient and accurate urban sound classification and recognition has become an important topic in the field of urban environmental monitoring. In this paper, we propose an urban noise classification method based on the Frequency Enhanced Convolution Attention (ConvNeXt-FECA) model. By fusing the Spectrogram and MFCC features in the early stage, the method makes full use of the advantages of the two features, and introduces the Frequency Enhanced Convolutional Attention Mechanism (FECA) to adaptively pay attention to the changes in the frequency band of the audio signal, which effectively improves the classification performance. Experimental results show that the classification accuracy of the ConvNeXt-FECA model is 98.5% on the Urban Sound8K dataset, showing strong robustness and generalization ability.
Downloads
References
[1] Chen X,Wang M,Kan R, et al. Improved Patch-Mix Transformer Sound Classification in Noisy Environments using Contrastive Learning methods [J]. Applied Sciences, 2024, 14 (21): 9711-9711.
[2] Talukder A M, Khalid M, Kazi M, et al. A hybrid cardiovascular arrhythmia disease detection using ConvNeXt-X models on electrocardiogram signals. [J]. Scientific reports, 2024, 14(1):30366.
[3] Huang Wenbo, Huang Yuxiang, Yao Yuan, et al. Automatic Grading of ConvNeXt Retinopathy Fused with Attention [J]. Optics and Precision Engineering, 2022,30(17):2147-2154.
[4] Wang J, Zhang B, Yin D, et al. Distribution network fault identification method based on multimodal ResNet with recorded waveform-driven feature extraction[J]. Energy Reports, 2025, 1390-104.
[5] Lozoya L S R ,DomÃnguez O J D H ,Azuela S H J , et al. Residual shallow convolutional neural network to classify microcalcifications clusters in digital mammograms[J]. Biomedical Signal Processing and Control,2025,102107209-107209.
[6] Schmidt A ,Gierden C, Heinen F R , et al.Efficient thermo-mechanically coupled and geometrically nonlinear two-scale FE-FFT-based modeling of elasto-viscoplastic polycrystalline materials[J].Computer Methods in Applied Mechanics and Engineering,2025,435117648-117648.
[7] Hantao Q, Xin G, Hualei X , et al.Comprehensive receptive field adaptive graph convolutional networks for action recognition[J].Journal of Visual Communication and Image Representation,2023,97.
[8] Chen Y , Zheng S , Wang H ,et al.ERes2NetV2: Boosting Short-Duration Speaker Verification Performance with Computational Efficiency[J]. arXiv preprint arXiv: 2406. 02167 (2024).
[9] Curtis J. KEEGAN ISSUES WARNING TO PLAYERS; FOOTBALL: Latest from the England Camp Plus Local Round-Up[J].[2024-12-25].
[10] Ning L, Weina J ,Xia L , et al.A high mechanical strength, deformable, fatigue-resistant polyacrylonitrile nanosphere-reinforced gel electrolyte for supercapacitors[J].Chemical Engineering Journal,2023,474.
[11] Yang Junjie, Ding Jiahui, Weng Shilong, et al. An automatic classification method for indoor environmental sounds based on lightweight ECAPA-TDNN neural network: CN2022117 15093.7[P].CN116013276A[2024-12-25].
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Frontiers in Computing and Intelligent Systems
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
