ICASSP 2020 results

Speaker independence of neural vocoders and their effect on parametric resynthesis speech enhancement [pdf]

Soumi Maiti and Michael Mandel

Traditional speech enhancement systems produce speech with compromised quality. Here we propose to use the high quality speech generation capability of neural vocoders for better quality speech enhancement. We term this parametric resynthesis (PR). In previous work, we showed that PR systems generate high quality speech for a single speaker using two neural vocoders, WaveNet and WaveGlow. Both these vocoders are traditionally speaker dependent. Here we first show that when trained on data from enough speakers, these vocoders can generate speech from unseen speakers, both male and female, with similar quality as seen speakers in training. Next using these two vocoders and a new vocoder LPCNet, we evaluate the noise reduction quality of PR on unseen speakers and show that objective signal and overall quality is higher than the state-of-the-art speech enhancement systems Wave-U-Net, Wavenet-denoise, and SEGAN. Moreover, in subjective quality, multiple-speaker PR out-performs the oracle Wiener mask.

Audio files

Wav files from the VCTK corpus mixed with noise [1] from the DEMAND dataset [2], as used in several previous papers [3-5].

References

[1] Cassia Valentini-Botinhao et al., "Noisy speech database for training speech enhancement algorithms and TTS models," University of Edinburgh. School of Informatics. Centre for Speech Technology Research (CSTR), 2017.

[2] Joachim Thiemann, Nobutaka Ito, and Emmanuel Vincent, "The diverse environments multi-channel acoustic noise database (DEMAND): A database of multichannel environmental noise recordings," in Proceedings of Meetings on Acoustics ICA2013. ASA, 2013, vol. 19, p. 035081.

[3] Santiago Pascual, Antonio Bonafonte, and Joan Serra, "SEGAN: Speech enhancement generative adversarial network," arXiv preprint arXiv:1703.09452, 2017.

[4] Dario Rethage, Jordi Pons, and Xavier Serra, "A wavenet forspeech denoising," in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, 2018, pp. 5069–5073.

[5] Craig Macartney and Tillman Weyde, "Improved speechenhancement with the wave-u-net," arXiv preprint arXiv:1811.11307, 2018.

Neural parametric resynthesis systems

FileNoisyPR-WorldPR-NVWaveNetPR-WaveGlowPR-LPCNetClean
p232_013
p232_054
p232_092
p232_104
p232_128
p232_169
p232_202
p232_302
p232_410
p257_011
p257_101
p257_157
p257_243
p257_365
p257_395

Baseline systems (from their released demos)

FileNoisySEGAN [3]Rethage et al [4]Wave-U-Net [5]Oracle WienerClean
p232_013
p232_054
p232_092
p232_104
p232_128
p232_169
p232_202
p232_302
p232_410
p257_011
p257_101
p257_157
p257_243
p257_365
p257_395