Distribution Augmentation for Low-Resource Expressive Text-To-Speech
By Mateusz Lajszczak, Animesh Prasad, Arent van Korlaar, Bajibabu Bollepalli, Antonio Bonafonte, Arnaud Joly, M. Nicolis, Alexis Moinet, Thomas Drugman, Trevor Wood and 1 more. 2022
10 citations recorded by the source corpus.
This paper presents a novel data augmentation technique for text-to-speech (TTS), that allows to generate new (text, audio) training examples without requiring any additional data. Our goal is to in-crease diversity of text conditionings available during training. This helps to reduce overfitting, especially in low-resource settings. Our method relies on substituting text and audio fragments in a way that preserves syntactical correctness. We take additional measures to ensure that synthesized speech does not contain artifacts caused by combining inconsistent audio samples. The perceptual evaluations show that our method improves speech quality over a number of datasets, speakers, and TTS architectures. We also demonstrate that it greatly improves robustness of attention-based TTS models.
Subfields: Artificial Intelligence, 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
- Antonio Bonafonte
- Björn Schuller
- Sri Karlapati
- Alexis Moinet
- Thomas Drugman
- Panagiotis Tzirakis
- Bartosz Putrycz
- Ruibo Fu
In Signal Processing
- Panagiotis Tzirakis
- Björn Schuller
- Antonio Bonafonte
- Alexis Moinet
- Thomas Drugman
- Andreas Triantafyllopoulos
- Bartosz Putrycz
- Ruibo Fu
Citations in the corpus
- BASE TTS: Lessons from building a billion-parameter Text-to-Speech model on 100K hours of data (2024, background): “Cars displayed in these showrooms with ENERGY-EFFICIENT AND GREEN decals prominently featured across the windshield aren’t announcing environmentalism; they’re virtue signaling.”
- An Overview of Affective Speech Synthesis and Conversion in the Deep Learning Era (2022, background): “Given the small sizes of these datasets, some works have experimented with data augmentation to account for the scarcity of data [178].”
- Controllable Emphasis with zero data for text-to-speech (2023, methodology): “The duration model as in [21, 27] consists of a stack of 3 convolution layers with 512 channels, kernel size of 5 and a dropout of 30%, a Bi-LSTM layer and a linear dense layer.”