Single-Channel Audio Source Separation Using Deep Neural Network Ensembles

Grais EM, Roma G, Simpson AJR, Plumbley MD. Single-Channel Audio Source Separation Using Deep Neural Network Ensembles. In: AES Convention 140. Audio Engineering Society; 2016. Paper 9494. Available from: https://aes.org/publications/elibrary-page/?id=18193

@inproceedings{Grais2016_18193,
  author = {Grais, Emad M. and Roma, Gerard and Simpson, Andrew J. R. and Plumbley, Mark D.},
  title = {{Single-Channel Audio Source Separation Using Deep Neural Network Ensembles}},
  booktitle = {AES Convention 140},
  note = {Paper 9494},
  year = {2016},
  month = may,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=18193}
}

TY  - CPAPER
TI  - Single-Channel Audio Source Separation Using Deep Neural Network Ensembles
AU  - Grais, Emad M.
AU  - Roma, Gerard
AU  - Simpson, Andrew J. R.
AU  - Plumbley, Mark D.
T2  - AES Convention 140
M1  - Paper 9494
PY  - 2016
DA  - 2016/05/06
UR  - https://aes.org/publications/elibrary-page/?id=18193
PB  - Audio Engineering Society
LA  - en
AB  - Deep neural networks (DNNs) are often used to tackle the single channel source separation (SCSS) problem by predicting time-frequency masks. The predicted masks are then used to separate the sources from the mixed signal. Different types of masks produce separated sources with different levels of distortion and interference. Some types of masks produce separated sources with low distortion, while other masks produce low interference between the separated sources. In this paper a combination of different DNNs’ predictions (masks) is used for SCSS to achieve better quality of the separated sources than using each DNN individually. We train four different DNNs by minimizing four different cost functions to predict four different masks. The first and second DNNs are trained to approximate reference binary and soft masks. The third DNN is trained to predict a mask from the reference sources directly. The last DNN is trained similarly to the third DNN but with an additional discriminative constraint to maximize the differences between the estimated sources. Our experimental results show that combining the predictions of different DNNs achieves separated sources with better quality than using each DNN individually.
ER  -