PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training

Close G, Hain T, Goetze S. PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training. In: AES Convention 154. Audio Engineering Society; 2023. Paper 10656. Available from: https://aes.org/publications/elibrary-page/?id=22063

@inproceedings{Close2023_22063,
  author = {Close, George and Hain, Thomas and Goetze, Stefan},
  title = {{PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training}},
  booktitle = {AES Convention 154},
  note = {Paper 10656},
  year = {2023},
  month = may,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=22063}
}

TY  - CPAPER
TI  - PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training
AU  - Close, George
AU  - Hain, Thomas
AU  - Goetze, Stefan
T2  - AES Convention 154
M1  - Paper 10656
PY  - 2023
DA  - 2023/05/06
UR  - https://aes.org/publications/elibrary-page/?id=22063
PB  - Audio Engineering Society
LA  - en
AB  - Recent speech enhancement work, which makes use of neural networks trained with a loss derived in part using an adversarial metric prediction network, has shown to be very effective. However, by limiting the data used to train this metric prediction network to only the clean reference and the output of the speech enhancement network, only a limited range of the metric is learnt. Additionally, such speech enhancement systems are limited because they typically operate solely over magnitude spectrogram representations so they do not encode phase information. In this work, recent developments for phase-aware speech enhancement in such an adversarial framework are expanded in two ways to enable the metric prediction network to learn a full range of metric scores. Firstly, the metric predictor is also exposed to unenhanced ’noisy’ data during training. Furthermore, an additional network is introduced and trained alongside which attempts to produce outputs with a fixed ’lower’ target metric score, and expose the metric predictor to these ’de-enhanced’ outputs. It is found that performance increases versus a baseline system utilising a magnitude spectrogram speech enhancement network.
ER  -