Voice Conversion with Conditional SampleRNN
This addresses voice conversion for applications requiring speaker identity changes without parallel data, representing a novel method for a known bottleneck.
The paper tackled voice conversion by conditioning SampleRNN on linguistic features, pitch contour, and speaker identity to preserve content and learn style, achieving many-to-many conversion without parallel data and outperforming conventional methods in subjective evaluations.
Here we present a novel approach to conditioning the SampleRNN generative model for voice conversion (VC). Conventional methods for VC modify the perceived speaker identity by converting between source and target acoustic features. Our approach focuses on preserving voice content and depends on the generative network to learn voice style. We first train a multi-speaker SampleRNN model conditioned on linguistic features, pitch contour, and speaker identity using a multi-speaker speech corpus. Voice-converted speech is generated using linguistic features and pitch contour extracted from the source speaker, and the target speaker identity. We demonstrate that our system is capable of many-to-many voice conversion without requiring parallel data, enabling broad applications. Subjective evaluation demonstrates that our approach outperforms conventional VC methods.