SpeechSplit 2.0: Unsupervised speech disentanglement for voice conversion Without tuning autoencoder Bottlenecks
This work addresses the challenge of time-consuming and less robust bottleneck tuning in voice conversion systems, offering a more efficient solution for researchers and practitioners in speech processing.
The paper tackled the problem of unsupervised speech disentanglement for voice conversion by proposing SpeechSplit 2.0, which replaces the need for tuning autoencoder bottlenecks with efficient signal processing methods, achieving comparable performance and superior robustness to bottleneck size variations.
SpeechSplit can perform aspect-specific voice conversion by disentangling speech into content, rhythm, pitch, and timbre using multiple autoencoders in an unsupervised manner. However, SpeechSplit requires careful tuning of the autoencoder bottlenecks, which can be time-consuming and less robust. This paper proposes SpeechSplit 2.0, which constrains the information flow of the speech component to be disentangled on the autoencoder input using efficient signal processing methods instead of bottleneck tuning. Evaluation results show that SpeechSplit 2.0 achieves comparable performance to SpeechSplit in speech disentanglement and superior robustness to the bottleneck size variations. Our code is available at https://github.com/biggytruck/SpeechSplit2.