Beyond Equal-Length Snippets: How Long Is Sufficient to Recognize an Audio Scene?

Phan H, Chén OY, Koch P, Pham L, McLoughlin I, Mertins A, De Vos M. Beyond Equal-Length Snippets: How Long Is Sufficient to Recognize an Audio Scene? In: AES Conference: 2019 AES International Conference on Audio Forensics. Audio Engineering Society; 2019. Paper 16. Available from: https://aes.org/publications/elibrary-page/?id=20468

@inproceedings{Phan2019_20468,
  author = {Phan, Huy and Chén, Oliver Y. and Koch, Philipp and Pham, Lam and McLoughlin, Ian and Mertins, Alfred and De Vos, Maarten},
  title = {{Beyond Equal-Length Snippets: How Long Is Sufficient to Recognize an Audio Scene?}},
  booktitle = {AES Conference: 2019 AES International Conference on Audio Forensics},
  note = {Paper 16},
  year = {2019},
  month = jun,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=20468}
}

TY  - CPAPER
TI  - Beyond Equal-Length Snippets: How Long Is Sufficient to Recognize an Audio Scene?
AU  - Phan, Huy
AU  - Chén, Oliver Y.
AU  - Koch, Philipp
AU  - Pham, Lam
AU  - McLoughlin, Ian
AU  - Mertins, Alfred
AU  - De Vos, Maarten
T2  - AES Conference: 2019 AES International Conference on Audio Forensics
M1  - Paper 16
PY  - 2019
DA  - 2019/06/06
UR  - https://aes.org/publications/elibrary-page/?id=20468
PB  - Audio Engineering Society
LA  - en
AB  - Due to the variability in characteristics of audio scenes, some scenes can naturally be recognized earlier than others. In this work, rather than using equal-length snippets for all scene categories, as is common in the literature, we study to which temporal extent an audio scene can be reliably recognized given state-of-the-art models. Moreover, as model fusion with deep network ensemble is prevalent in audio scene classi?cation, we further study whether, and if so, when model fusion is necessary for this task. To achieve these goals, we employ two single-network systems relying on a convolutional neural network and a recurrent neural network for classi?cation as well as early fusion and late fusion of these networks. Experimental results on the LITIS-Rouen dataset show that some scenes can be reliably recognized with a few seconds while other scenes require signi?cantly longer durations. In addition, model fusion is shown to be the most bene?cial when the signal length is short.
ER  -