Alert button

Voice conversion with limited data and limitless data augmentations

Dec 27, 2022
Olga Slizovskaia, Jordi Janer, Pritish Chandna, Oscar Mayor

Figure 1 for Voice conversion with limited data and limitless data augmentations
Figure 2 for Voice conversion with limited data and limitless data augmentations
Figure 3 for Voice conversion with limited data and limitless data augmentations
Figure 4 for Voice conversion with limited data and limitless data augmentations

Share this with someone who'll enjoy it:

Applying changes to an input speech signal to change the perceived speaker of speech to a target while maintaining the content of the input is a challenging but interesting task known as Voice conversion (VC). Over the last few years, this task has gained significant interest where most systems use data-driven machine learning models. Doing the conversion in a low-latency real-world scenario is even more challenging constrained by the availability of high-quality data. Data augmentations such as pitch shifting and noise addition are often used to increase the amount of data used for training machine learning based models for this task. In this paper we explore the efficacy of common data augmentation techniques for real-time voice conversion and introduce novel techniques for data augmentation based on audio and voice transformation effects as well. We evaluate the conversions for both male and female target speakers using objective and subjective evaluation methodologies.

View paper onarxiv icon

Share this with someone who'll enjoy it: