Bias in Automated Speaker Recognition

Bias in Automated Speaker Recognition

Hutiri, Wiebke (TU Delft Information and Communication Technology)
Ding, Aaron Yi (TU Delft Information and Communication Technology)

2022

Automated speaker recognition uses data processing to identify speakers by their voice. Today, automated speaker recognition is deployed on billions of smart devices and in services such as call centres. Despite their wide-scale deployment and known sources of bias in related domains like face recognition and natural language processing, bias in automated speaker recognition has not been studied systematically. We present an in-depth empirical and analytical study of bias in the machine learning development workflow of speaker verification, a voice biometric and core task in automated speaker recognition. Drawing on an established framework for understanding sources of harm in machine learning, we show that bias exists at every development stage in the well-known VoxCeleb Speaker Recognition Challenge, including data generation, model building, and implementation. Most affected are female speakers and non-US nationalities, who experience significant performance degradation. Leveraging the insights from our findings, we make practical recommendations for mitigating bias in automated speaker recognition, and outline future research directions.

audit
bias
evaluation
fairness
speaker recognition
speaker verification

http://resolver.tudelft.nl/uuid:a6adeb57-3211-45ad-9329-9f134c82d790

https://doi.org/10.1145/3531146.3533089

Association for Computing Machinery (ACM)

978-1-4503-9352-2

Proceedings of 2022 5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022

5th ACM Conference on Fairness, Accountability, and Transparency, FAccT 2022, 2022-06-21 → 2022-06-24, Virtual, Online, Korea, Republic of

ACM International Conference Proceeding Series

Institutional Repository

conference paper

© 2022 Wiebke Hutiri, Aaron Yi Ding