Hierarchical disentangled representation learning for singing voice conversion
This work addresses a specific bottleneck in singing voice conversion for audio processing applications, representing an incremental improvement.
The paper tackles the problem of high-dimensional data in singing voice conversion by proposing a hierarchical representation learning method that disentangles multiple resolutions, resulting in improved mean opinion score, similarity score, and pitch accuracy compared to single-resolution baselines.
Conventional singing voice conversion (SVC) methods often suffer from operating in high-resolution audio owing to a high dimensionality of data. In this paper, we propose a hierarchical representation learning that enables the learning of disentangled representations with multiple resolutions independently. With the learned disentangled representations, the proposed method progressively performs SVC from low to high resolutions. Experimental results show that the proposed method outperforms baselines that operate with a single resolution in terms of mean opinion score (MOS), similarity score, and pitch accuracy.