Measuring the Effectiveness of Voice Conversion on Speaker Identification and Automatic Speech Recognition Systems

TL;DR

This study assesses a Cycle-GAN based voice converter's ability to mimic speakers and its impact on speech recognition systems, finding limited benefits for ASR training despite successful speaker imitation.

Contribution

It evaluates the effectiveness of a Cycle-GAN voice converter on speaker ID and ASR systems and explores its potential to enhance ASR training with synthetic data.

Findings

Voice converter achieves up to 46% top-1 accuracy in speaker identification.

Synthetic data marginally improves speech recognition error rates.

Voice imitation is successful, but benefits for ASR training are limited.

Abstract

This paper evaluates the effectiveness of a Cycle-GAN based voice converter (VC) on four speaker identification (SID) systems and an automated speech recognition (ASR) system for various purposes. Audio samples converted by the VC model are classified by the SID systems as the intended target at up to 46% top-1 accuracy among more than 250 speakers. This encouraging result in imitating the target styles led us to investigate if converted (synthetic) samples can be used to improve ASR training. Unfortunately, adding synthetic data to the ASR training set only marginally improves word and character error rates. Our results indicate that even though VC models can successfully mimic the style of target speakers as measured by SID systems, improving ASR training with synthetic data from VC systems needs further research to establish its efficacy.