PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training
×
Cite This
Citation & Abstract
G. Close, T. Hain, and S. Goetze, "PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training," Paper 10656, (2023 May.). doi:
G. Close, T. Hain, and S. Goetze, "PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training," Paper 10656, (2023 May.). doi:
Abstract: 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.
@article{close2023pamgan+/-:,
author={close, george and hain, thomas and goetze, stefan},
journal={journal of the audio engineering society},
title={pamgan+/-: improving phase-aware speech enhancement performance via expanded discriminator training},
year={2023},
volume={},
number={},
pages={},
doi={},
month={may},}
@article{close2023pamgan+/-:,
author={close, george and hain, thomas and goetze, stefan},
journal={journal of the audio engineering society},
title={pamgan+/-: improving phase-aware speech enhancement performance via expanded discriminator training},
year={2023},
volume={},
number={},
pages={},
doi={},
month={may},
abstract={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.},}
TY - Capturing
TI - PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training
SP -
EP -
AU - Close, George
AU - Hain, Thomas
AU - Goetze, Stefan
PY - 2023
JO - Journal of the Audio Engineering Society
IS -
VO -
VL -
Y1 - May 2023
TY - Capturing
TI - PAMGAN+/-: Improving Phase-Aware Speech Enhancement Performance via Expanded Discriminator Training
SP -
EP -
AU - Close, George
AU - Hain, Thomas
AU - Goetze, Stefan
PY - 2023
JO - Journal of the Audio Engineering Society
IS -
VO -
VL -
Y1 - May 2023
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.
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.
Open Access
Authors:
Close, George; Hain, Thomas; Goetze, Stefan
Affiliations:
The University of Sheffield, UK; The University of Sheffield, UK; The University of Sheffield, UK(See document for exact affiliation information.)
AES Convention:
154 (May 2023)
Paper Number:
10656
Publication Date:
May 13, 2023Import into BibTeX
Subject:
Capturing
Permalink:
http://www.aes.org/e-lib/browse.cfm?elib=22063