Investigating RNN-based speech enhancement methods for noise-robust Text-to-Speech

By Cassia Valentini-Botinhao, Xin Eric Wang, Shinji Takaki, Junichi Yamagishi. 2016

554 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xu Tan
2. Tao Qin
3. Tie-Yan Liu
4. Yi Ren
5. Sheng Zhao
6. Jinglin Liu
7. Shinji Watanabe
8. Junichi Yamagishi

In Speech & Audio

1. Xu Tan
2. Tao Qin
3. Tie-Yan Liu
4. Sheng Zhao
5. Yi Ren
6. Jinglin Liu
7. Junichi Yamagishi
8. Nima Mesgarani

In Signal Processing

1. Xu Tan
2. Tao Qin
3. Tie-Yan Liu
4. Deliang Wang
5. Yi Ren
6. Nima Mesgarani
7. Jinglin Liu
8. Shinji Watanabe

Citations in the corpus

• ELLA-V: Stable Neural Codec Language Modeling with Alignment-guided Sequence Reordering (2024, methodology): “With the advancement of modern neural networks, some work initially replaced HMMs with recurrent neural networks (RNNs) but still followed the SPSS paradigm (Fan et al., 2014; Zen and Sak, 2015; Valentini-Botinhao et al., 2016).”
• Speech Denoising with Deep Feature Losses (2018, background): “All details concerning the data can be found in [41].”
• A Wavenet for Speech Denoising (2017, methodology): “The dataset we used [12, 18] was generated from two sources: speech data was supplied by the Voice Bank corpus [19] while environmental sounds were provided by the Diverse Environments Multichannel Acoustic Noise Database (DEMAND) [20].”
• SEGAN: Speech Enhancement Generative Adversarial Network (2017, methodology): “To make the noisy training set, a total of 40 different conditions are considered [27]: 10 types of noise (2 artificial and 8 from the Demand database [29]) with 4 signal-to-noise ratio (SNR) each (15, 10, 5, and 0dB).”
• Noise Robust TTS for Low Resource Speakers using Pre-trained Model and Speech Enhancement (2020, methodology): “Authors in [9] introduced speech enhancement methods for noise robust TTS.”
• Speech Enhancement of Noisy and Reverberant Speech for Text-to-Speech (2018, methodology): “In this article we revise parts of the work reported in [25]: the construction of a noisy version of an existing database that was purposely created for training TTS voices and the evaluation of RNN-based enhancement models on this data.”
• MetricGAN+: An Improved Version of MetricGAN for Speech Enhancement (2021, methodology): “To compare the proposed MetricGAN+ with other existing methods, we use the publicly available VoiceBank-DEMAND dataset [27].”
• Denoispeech: Denoising Text to Speech with Frame-Level Noise Modeling (2020, methodology): “A straight-forward way [1, 2, 3, 4] is to first de-noise the noisy speech with a pre-trained speech enhancement model [5], and then use the enhanced speech to train the TTS model.”
• Interactive Speech and Noise Modeling for Speech Enhancement (2020, methodology): “(Valentini-Botinhao et al. 2016).”
• Speech Enhancement and Dereverberation With Diffusion-Based Generative Models (2022, unknown): “b) VB-DMD: We use the publicly available VoiceBankDEMAND dataset (VB-DMD) [53] which is often used as a benchmark for single-channel speech enhancement.”
• FRCRN: Boosting Feature Representation Using Frequency Recurrence for Monaural Speech Enhancement (2022, unknown): “We further conducted evaluations on two popular benchmarks: DNS-2020 dataset [11] and VoiceBank+Demand dataset [12] 2 and 3 show the performance comparisons with the published results of the SOTA models.”
• PHASEN: A Phase-and-Harmonics-Aware Speech Enhancement Network (2019, background): “Voice Bank+DEMAND: This is an open dataset1 proposed by (Valentini-Botinhao et al. 2016).”
• Conditional Diffusion Probabilistic Model for Speech Enhancement (2022, methodology): “Dataset: we evaluate the CDiffuSE model on the VoiceBankDEMAND dataset [26].”
• Improving GANs for Speech Enhancement (2020, methodology): “To assess the performance of the proposed ISEGAN and DSEGAN and demonstrate their advantages over the vanilla SEGAN, we carried out experiments on the database in [27] on which the vanilla SEGAN was evaluated [14].”
• NADiffuSE: Noise-aware Diffusion-based Model for Speech Enhancement (2023, methodology): “The new one follows the same setting which consists of 30 speakers from the VoiceBank [40] corpus and was mixed with 12”
• Sergan: Speech Enhancement Using Relativistic Generative Adversarial Networks with Gradient Penalty (2019, methodology): “We used the dataset presented in [22] for comparing the various systems.”
• MetricGAN: Generative Adversarial Networks based Black-box Metric Scores Optimization for Speech Enhancement (2019, methodology): “To further compare the proposed MetricGAN with other state-of-the-art methods, we use a publicly available dataset released by (Valentini-Botinhao et al., 2016).”