Effects of Convolutional Autoencoder Bottleneck Width on StarGAN-based Singing Technique Conversion

TL;DR

This study investigates how the width of the convolutional autoencoder bottleneck influences the quality of singing technique conversion using a GAN-based system, with findings on articulation clarity and conversion ease.

Contribution

It introduces a systematic evaluation of CAE bottleneck widths in a GAN-based singing technique conversion system, highlighting their effects on synthesis quality and technique transfer.

Findings

Wider bottlenecks improve articulation clarity.

Wider bottlenecks do not necessarily increase target technique likeness.

Whistle voice is the easiest target for conversion.

Abstract

Singing technique conversion (STC) refers to the task of converting from one voice technique to another while leaving the original singer identity, melody, and linguistic components intact. Previous STC studies, as well as singing voice conversion research in general, have utilized convolutional autoencoders (CAEs) for conversion, but how the bottleneck width of the CAE affects the synthesis quality has not been thoroughly evaluated. To this end, we constructed a GAN-based multi-domain STC system which took advantage of the WORLD vocoder representation and the CAE architecture. We varied the bottleneck width of the CAE, and evaluated the conversion results subjectively. The model was trained on a Mandarin dataset which features four singers and four singing techniques: the chest voice, the falsetto, the raspy voice, and the whistle voice. The results show that a wider bottleneck corresponds to better articulation clarity but does not necessarily lead to higher likeness to the target technique. Among the four techniques, we also found that the whistle voice is the easiest target for conversion, while the other three techniques as a source produce more convincing conversion results than the whistle.