iMetricGAN: Intelligibility Enhancement for Speech-in-Noise Using Generative Adversarial Network-Based Metric Learning

Haoyu Li, Szu-Wei Fu, Yu Tsao, Junichi Yamagishi


The intelligibility of natural speech is seriously degraded when exposed to adverse noisy environments. In this work, we propose a deep learning-based speech modification method to compensate for the intelligibility loss, with the constraint that the root mean square (RMS) level and duration of the speech signal are maintained before and after modifications. Specifically, we utilize an iMetricGAN approach to optimize the speech intelligibility metrics with generative adversarial networks (GANs). Experimental results show that the proposed iMetricGAN outperforms conventional state-of-the-art algorithms in terms of objective measures, i.e., speech intelligibility in bits (SIIB) and extended short-time objective intelligibility (ESTOI), under a Cafeteria noise condition. In addition, formal listening tests reveal significant intelligibility gains when both noise and reverberation exist.


 DOI: 10.21437/Interspeech.2020-1016

Cite as: Li, H., Fu, S., Tsao, Y., Yamagishi, J. (2020) iMetricGAN: Intelligibility Enhancement for Speech-in-Noise Using Generative Adversarial Network-Based Metric Learning. Proc. Interspeech 2020, 1336-1340, DOI: 10.21437/Interspeech.2020-1016.


@inproceedings{Li2020,
  author={Haoyu Li and Szu-Wei Fu and Yu Tsao and Junichi Yamagishi},
  title={{iMetricGAN: Intelligibility Enhancement for Speech-in-Noise Using Generative Adversarial Network-Based Metric Learning}},
  year=2020,
  booktitle={Proc. Interspeech 2020},
  pages={1336--1340},
  doi={10.21437/Interspeech.2020-1016},
  url={http://dx.doi.org/10.21437/Interspeech.2020-1016}
}