Training Multi-Speaker Neural Text-to-Speech Systems using Speaker-Imbalanced Speech Corpora
This addresses a practical data imbalance issue in TTS systems for applications requiring diverse speaker voices, though it is incremental in nature.
The paper tackles the problem of training multi-speaker neural text-to-speech systems with imbalanced speaker data, finding that simply combining all data yields performance similar to or better than speaker-dependent models, and using an ensemble model further improves speech quality for underrepresented speakers.
When the available data of a target speaker is insufficient to train a high quality speaker-dependent neural text-to-speech (TTS) system, we can combine data from multiple speakers and train a multi-speaker TTS model instead. Many studies have shown that neural multi-speaker TTS model trained with a small amount data from multiple speakers combined can generate synthetic speech with better quality and stability than a speaker-dependent one. However when the amount of data from each speaker is highly unbalanced, the best approach to make use of the excessive data remains unknown. Our experiments showed that simply combining all available data from every speaker to train a multi-speaker model produces better than or at least similar performance to its speaker-dependent counterpart. Moreover by using an ensemble multi-speaker model, in which each subsystem is trained on a subset of available data, we can further improve the quality of the synthetic speech especially for underrepresented speakers whose training data is limited.