Neural Text-to-Speech Adaptation from Low Quality Public Recordings

By Qiong Hu, Erik Marchi, David Winarsky, Yannis Stylianou, Devang Naik, Sachin S. Kajarekar. 2019

27 citations recorded by the source corpus.

Neural Text-to-Speech (TTS) synthesis is able to generate high-quality speech with natural prosody. However, these systems typically require a large amount of data, preferably recorded in a clean and noise-free environment. We focus on creating target voices from low quality public recordings and our ﬁndings show that even with a large amount of data from a speciﬁc speaker, it is challenging to train a speaker-dependent neural TTS model. In order to improve the voice quality, while simultaneously reducing the amount of data required, we introduce meta-learning to adapt the neural TTS front-end. We propose three approaches for multi-speaker systems: (1) a lookup-table-based system, (2) a speaker representation derived from the Personalized Hey Siri (PHS) system, and (3) a system with no speaker encoder. Re-sults show that: i) By using a signiﬁcantly smaller number of target voice recordings, the proposed system based on embed-dings trained from the PHS system can generate comparable quality and speaker similarity to the speaker-dependent model trained solely on the target voice. ii) Applying meta-learning to Tacotron can effectively learn a representation of an unseen speaker. iii) For low quality public recordings, the adaptation based on the multi-speaker corpus can generate a cleaner target voice in comparison with the speaker-dependent model.

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

In Signal Processing

Citations in the corpus

A Survey on Neural Speech Synthesis (2021, methodology): “[119] propose to mine this kind of data and develop sophisticated techniques to train a TTS model.”
Zero-Shot Multi-Speaker Text-To-Speech with State-Of-The-Art Neural Speaker Embeddings (2019, methodology): “Prior studies have focused on training a speaker encoder network jointly with the TTS model [6, 8, 9] or the neural vocoder [10]; others have explored the use of speaker embeddings in combination with fine-tuning the TTS model [10, 11, 12].”
Can Speaker Augmentation Improve Multi-Speaker End-to-End TTS? (2020, background): “In [10], low-quality recordings were used for fine-tuning based speaker adaptation.”
Data Efficient Voice Cloning from Noisy Samples with Domain Adversarial Training (2020, background): “In order to build a robust TTS system, there are several attempts to conduct speech synthesis with noisy data [10, 11, 12].”
Generating Multilingual Voices Using Speaker Space Translation Based on Bilingual Speaker Data (2020, background): “In [18] it was found useful for inlanguage voice adaptation.”
Speaker Conditional WaveRNN: Towards Universal Neural Vocoder for Unseen Speaker and Recording Conditions (2020, methodology): “Fine-tuning pretrained TTS model in combination with speaker embeddings was addressed in [31, 32, 33].”
Attentron: Few-Shot Text-to-Speech Utilizing Attention-Based Variable-Length Embedding (2020, unknown): “To avoid it, some approaches [4, 10, 11, 12, 13, 14, 20] jointly or individually trained a speaker encoder generating a global embedding and utilized it to condition the TTS model.”
Text-Free Image-to-Speech Synthesis Using Learned Segmental Units (2020, background): “by a synthesis model benefits from training data recorded by a single speaker in a clean recording environment [53, 110, 29, 52].”
Effective and Differentiated Use of Control Information for Multi-speaker Speech Synthesis (2021, background): “It can be further adapted in different tasks such as Speaker adaptation [7], cross-lingual Text-to-speech synthesis [8], and style conversion [9].”