Abstract
Audio classification aims to distinguish different kinds of sounds, and it is of great importance to artificial intelligence applications. Nevertheless, there are still some challenges faced in this field, especially the classification of weakly labeled audio signals. The audio clip contains temporal information and spatial information. However, existing methods only utilize partial information so that the classification effect requires to be improved. To improve classification accuracy, we propose a multi-level attention fusion network (MLAFNet) based on deep supervision which includes multi-attention fusion (MAF) module and multi-level fusion (MLF) module. The MAF module can take full advantage of the information from the time and space domain. The MLF module based on deep supervision strategy can combine the coarse-grained and fine-grained information. Extensive experiments are carried out on the basis of Google Audio Set to demonstrate the effectiveness of the proposed network beyond several state-of-the-art approaches, which achieve 0.970 on AUC and 2.652 on d-prime.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Mesaros, A., Heittola, T., Diment, A., Elizalde, B., Shah, A., Vincent, E., Raj, B., Virtanen, T.: Dcase 2017 challenge setup: Tasks, datasets and baseline system (2017)
Gemmeke, J.F., Ellis, D.P.W., Freedman, D., Jansen, A., Lawrence, W., Moore, R.C., Plakal, M., Ritter, M.: Audio set: An ontology and human-labeled dataset for audio events. In: 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 776–780. IEEE (2017)
Hershey, S., Chaudhuri, S., Ellis, D.P.W., Gemmeke, J.F., Jansen, A., Moore, R.C., Plakal, M., Platt, D., Saurous, R.A., Seybold, B., et al.: CNN architectures for large-scale audio classification. In: 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 131–135. IEEE (2017)
Maron, O., Lozano-P´erez, T.: A framework for multiple-instance learning. In: Advances in Neural Information Processing Systems, pp. 570–576 (1998)
Ilse, M., Tomczak, J.M., Welling, M.: Attention-based deep multiple instance learning. arXiv preprint arXiv:1802.04712 (2018)
Kong, Q., Xu, Y., Wang, W., Plumbley, M.D.: Audio set classification with attention model: a probabilistic perspective. In: 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 316–320. IEEE (2018)
Kong, Q., Yu, C., Xu, Y., Iqbal, T., Wang, W., Plumbley, M.D.: Weakly labelled audioset tagging with attention neural networks. IEEE/ACM Trans. Audio, Speech, Lang. Process., 27(11), 1791–1802 (2019)
Adavanne, S., Virtanen, T.: Sound event detection using weakly labeled dataset with stacked convolutional and recurrent neural network. arXiv preprint arXiv:1710.02998 (2017)
Yin, Y., Shah, R.R., Zimmermann, R.: Learning and fusing multimodal deep features for acoustic scene categorization. In: 2018 ACM Multimedia Conference on Multimedia Conference, pp. 1892–1900. ACM (2018)
Yu, C., Barsim, K.S., Kong, Q., Yang, B.: Multi-level attention model for weakly supervised audio classification. arXiv preprint arXiv:1803.02353 (2018)
Lee, C.-Y., Xie, S., Gallagher, P., Zhang, Z., Tu, Z.: Deeply-supervised nets. In: Artificial Intelligence and Statistics, pp. 562–570 (2015)
Wang, Y., Li, J., Metze, F.: A comparison of five multiple instance learning pooling functions for sound event detection with weak labeling. In: ICASSP 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 31–35. IEEE (2019)
Acknowledgment
This research was funded by the BUPT Basic Research Funding 500419757 and the National Natural Science Foundation of China under Grant 61901049.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Zhao, W., He, Y., Mu, J., Jing, X. (2021). A Multi-level Attention Fusion Network for Weakly Supervised Audio Classification. In: Wang, Y., Xu, L., Yan, Y., Zou, J. (eds) Signal and Information Processing, Networking and Computers. Lecture Notes in Electrical Engineering, vol 677. Springer, Singapore. https://doi.org/10.1007/978-981-33-4102-9_83
Download citation
DOI: https://doi.org/10.1007/978-981-33-4102-9_83
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-33-4101-2
Online ISBN: 978-981-33-4102-9
eBook Packages: EngineeringEngineering (R0)