VoxSim: A perceptual voice similarity dataset

Junseok Ahn1, Youkyum Kim1, Yeunju Choi2, Doyeop Kwak1, Ji-Hoon Kim1, Seongkyu Mun2, Joon Son Chung1

1Korea Advanced Institute of Science and Technology, South Korea

2Samsung Research, South Korea

Paper Code

Abstract

This paper introduces VoxSim, a dataset of perceptual voice similarity ratings. Recent efforts to automate the assessment of speech synthesis technologies have primarily focused on predicting mean opinion score of naturalness, leaving speaker voice similarity relatively unexplored due to a lack of extensive training data. To address this, we generate about 41k utterance pairs from the VoxCeleb dataset, a widely utilised speech dataset for speaker recognition, and collect nearly 70k speaker similarity scores through a listening test. VoxSim offers a valuable resource for the development and benchmarking of speaker similarity prediction models. We provide baseline results of speaker similarity prediction models on the VoxSim test set and further demonstrate the model trained on our dataset generalises to the out-of-domain VCC2018 dataset.

Overview

VoxSim is a dataset contatining perceptual voice similarity rating of VoxCeleb1 speech clips. There is no speaker overlap between the clips used to create the test set and train set. The dataset statistics are given in the table below.

Set # speakers # speaker combinations # pairs # ratings
Train 1,142 24,764 38,802 63,845
Test 109 904 2,776 5,564
Total 1,251 25,668 41,578 69,409

Downloads

Train List (raw score)
Train List (average score per utterance pair)
Test List (csv)

Each line represents {clip1_path},{clip2_path},{same_or_different_speaker},{listener_id},{rating}.

To download the VoxCeleb1 dataset, please refer to here.

License

The VoxSim dataset is available to download for research purposes under a Creative Commons Attribution 4.0 International License. A complete version of the license can be found here.

BibTex

@inproceedings{voxsim,
  author={Junseok Ahn, Youkyum Kim, Yeunju Choi, Doyeop Kwak, Ji-Hoon Kim, Seongkyu Mun, Joon Son Chung},
  booktitle={Proc. Interspeech 2024}, 
  title={VoxSim: A perceptual voice similarity dataset}, 
  year={2024},
}

Acknowledgement

This work was supported by Samsung Research.