The generative adversarial network (GAN) has shown its outstanding capability in improving Non-Autoregressive TTS (NAR-TTS) by adversarially training it with an extra model that discriminates between the real and the generated speech. To maximize the benefits of GAN, it is crucial to find a powerful discriminator that can capture rich distinguishable information. In this paper, we propose a multi-scale time-frequency spectrogram discriminator to help NAR-TTS generate high-fidelity Mel-spectrograms. It treats the spectrogram as a 2D image to exploit the correlation among different components in the time-frequency domain. And a U-Net-based model structure is employed to discriminate at different scales to capture both coarse-grained and fine-grained information. We conduct subjective tests to evaluate the proposed approach. Both multi-scale and time-frequency discriminating bring significant improvement in the naturalness and fidelity. When combining the neural vocoder, it is shown more effective and concise than fine-tuning the vocoder. Finally, we visualize the discriminating maps to compare their difference to verify the effectiveness of multi-scale discriminating.
Cite as: Guo, H., Lu, H., Wu, X., Meng, H. (2022) A Multi-Scale Time-Frequency Spectrogram Discriminator for GAN-based Non-Autoregressive TTS. Proc. Interspeech 2022, 1566-1570, doi: 10.21437/Interspeech.2022-52
@inproceedings{guo22_interspeech,
author={Haohan Guo and Hui Lu and Xixin Wu and Helen Meng},
title={{A Multi-Scale Time-Frequency Spectrogram Discriminator for GAN-based Non-Autoregressive TTS}},
year=2022,
booktitle={Proc. Interspeech 2022},
pages={1566--1570},
doi={10.21437/Interspeech.2022-52}
}