Combining Fully Convolutional and Recurrent Neural Networks for Single Channel Audio Source Separation

Grais EM, Plumbley MD. Combining Fully Convolutional and Recurrent Neural Networks for Single Channel Audio Source Separation. In: AES Convention 144. Audio Engineering Society; 2018. Paper 9990. Available from: https://aes.org/publications/elibrary-page/?id=19507

@inproceedings{Grais2018_19507,
  author = {Grais, Emad M. and Plumbley, Mark D.},
  title = {{Combining Fully Convolutional and Recurrent Neural Networks for Single Channel Audio Source Separation}},
  booktitle = {AES Convention 144},
  note = {Paper 9990},
  year = {2018},
  month = may,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=19507}
}

TY  - CPAPER
TI  - Combining Fully Convolutional and Recurrent Neural Networks for Single Channel Audio Source Separation
AU  - Grais, Emad M.
AU  - Plumbley, Mark D.
T2  - AES Convention 144
M1  - Paper 9990
PY  - 2018
DA  - 2018/05/06
UR  - https://aes.org/publications/elibrary-page/?id=19507
PB  - Audio Engineering Society
LA  - en
AB  - Combining different models is a common strategy to build a good audio source separation system. In this work we combine two powerful deep neural networks for audio single channel source separation (SCSS). Namely, we combine fully convolutional neural networks (FCNs) and recurrent neural networks, specifically, bidirectional long short-term memory recurrent neural networks (BLSTMs). FCNs are good at extracting useful features from the audio data and BLSTMs are good at modeling the temporal structure of the audio signals. Our experimental results show that combining FCNs and BLSTMs achieves better separation performance than using each model individually.
ER  -