An Improved StarGAN for Emotional Voice Conversion: Enhancing Voice Quality and Data Augmentation

TL;DR

This paper introduces an improved StarGAN framework with a two-stage training process that better disentangles emotional features from content, resulting in higher quality emotional voice conversion and improved data augmentation for speech emotion recognition.

Contribution

The paper presents a novel StarGAN model with a two-stage training process that effectively separates emotional features, reducing audio distortion and enhancing data augmentation capabilities.

Findings

Significantly reduced distortion in emotional voice conversion.

Achieved 2% higher Micro-F1 and 5% higher Macro-F1 in emotion recognition.

Outperformed baseline StarGAN in both objective and subjective evaluations.

Abstract

Emotional Voice Conversion (EVC) aims to convert the emotional style of a source speech signal to a target style while preserving its content and speaker identity information. Previous emotional conversion studies do not disentangle emotional information from emotion-independent information that should be preserved, thus transforming it all in a monolithic manner and generating audio of low quality, with linguistic distortions. To address this distortion problem, we propose a novel StarGAN framework along with a two-stage training process that separates emotional features from those independent of emotion by using an autoencoder with two encoders as the generator of the Generative Adversarial Network (GAN). The proposed model achieves favourable results in both the objective evaluation and the subjective evaluation in terms of distortion, which reveals that the proposed model can effectively reduce distortion. Furthermore, in data augmentation experiments for end-to-end speech emotion recognition, the proposed StarGAN model achieves an increase of 2% in Micro-F1 and 5% in Macro-F1 compared to the baseline StarGAN model, which indicates that the proposed model is more valuable for data augmentation.