Exploiting Deep Neural Networks for Two-to-Five Channel Surround Decoder
×
Cite This
Citation & Abstract
J. Choi, and J. Chang, "Exploiting Deep Neural Networks for Two-to-Five Channel Surround Decoder," J. Audio Eng. Soc., vol. 68, no. 12, pp. 938-949, (2020 December.). doi: https://doi.org/10.17743/jaes.2020.0020
J. Choi, and J. Chang, "Exploiting Deep Neural Networks for Two-to-Five Channel Surround Decoder," J. Audio Eng. Soc., vol. 68 Issue 12 pp. 938-949, (2020 December.). doi: https://doi.org/10.17743/jaes.2020.0020
Abstract: We exploited deep neural networks (DNN) for two-to-five channel surround decoding. Specifically DNNs are used to replace the primary-ambient separation and ambient-signal-rendering modules. For the training, the mean-squared error of the magnitude spectra between the decoded and five-channel target signals and the interchannel level differences between the target signals were used as the loss functions. Through this procedure the DNNs can derive the spectral weights that can be used to produce the decoded signals, similar to that for the target signals. The log spectral distance, signal-to-distortion ratio, and multiple stimuli with hidden reference and anchor tests were used for objective and subjective evaluations. The experimental results show that exploiting the DNNs can generate decoded signals that are more similar to the target signals than those obtained via previous methods.
@article{choi2021exploiting,
author={choi, jeonghwan and chang, joon-hyuk},
journal={journal of the audio engineering society},
title={exploiting deep neural networks for two-to-five channel surround decoder},
year={2021},
volume={68},
number={12},
pages={938-949},
doi={https://doi.org/10.17743/jaes.2020.0020},
month={december},}
@article{choi2021exploiting,
author={choi, jeonghwan and chang, joon-hyuk},
journal={journal of the audio engineering society},
title={exploiting deep neural networks for two-to-five channel surround decoder},
year={2021},
volume={68},
number={12},
pages={938-949},
doi={https://doi.org/10.17743/jaes.2020.0020},
month={december},
abstract={we exploited deep neural networks (dnn) for two-to-five channel surround decoding. specifically dnns are used to replace the primary-ambient separation and ambient-signal-rendering modules. for the training, the mean-squared error of the magnitude spectra between the decoded and five-channel target signals and the interchannel level differences between the target signals were used as the loss functions. through this procedure the dnns can derive the spectral weights that can be used to produce the decoded signals, similar to that for the target signals. the log spectral distance, signal-to-distortion ratio, and multiple stimuli with hidden reference and anchor tests were used for objective and subjective evaluations. the experimental results show that exploiting the dnns can generate decoded signals that are more similar to the target signals than those obtained via previous methods.},}
TY - paper
TI - Exploiting Deep Neural Networks for Two-to-Five Channel Surround Decoder
SP - 938
EP - 949
AU - Choi, Jeonghwan
AU - Chang, Joon-Hyuk
PY - 2021
JO - Journal of the Audio Engineering Society
IS - 12
VO - 68
VL - 68
Y1 - December 2020
TY - paper
TI - Exploiting Deep Neural Networks for Two-to-Five Channel Surround Decoder
SP - 938
EP - 949
AU - Choi, Jeonghwan
AU - Chang, Joon-Hyuk
PY - 2021
JO - Journal of the Audio Engineering Society
IS - 12
VO - 68
VL - 68
Y1 - December 2020
AB - We exploited deep neural networks (DNN) for two-to-five channel surround decoding. Specifically DNNs are used to replace the primary-ambient separation and ambient-signal-rendering modules. For the training, the mean-squared error of the magnitude spectra between the decoded and five-channel target signals and the interchannel level differences between the target signals were used as the loss functions. Through this procedure the DNNs can derive the spectral weights that can be used to produce the decoded signals, similar to that for the target signals. The log spectral distance, signal-to-distortion ratio, and multiple stimuli with hidden reference and anchor tests were used for objective and subjective evaluations. The experimental results show that exploiting the DNNs can generate decoded signals that are more similar to the target signals than those obtained via previous methods.
We exploited deep neural networks (DNN) for two-to-five channel surround decoding. Specifically DNNs are used to replace the primary-ambient separation and ambient-signal-rendering modules. For the training, the mean-squared error of the magnitude spectra between the decoded and five-channel target signals and the interchannel level differences between the target signals were used as the loss functions. Through this procedure the DNNs can derive the spectral weights that can be used to produce the decoded signals, similar to that for the target signals. The log spectral distance, signal-to-distortion ratio, and multiple stimuli with hidden reference and anchor tests were used for objective and subjective evaluations. The experimental results show that exploiting the DNNs can generate decoded signals that are more similar to the target signals than those obtained via previous methods.