Multi-Stage Face-Voice Association Learning with Keynote Speaker Diarization

TL;DR

This paper introduces MFV-KSD, a multi-stage framework for cross-modal face-voice association that effectively handles noisy inputs and improves inter-modal correlation, achieving top performance in a multilingual challenge.

Contribution

The paper proposes a novel three-stage training strategy and a keynote speaker diarization front-end for robust face-voice association learning.

Findings

Achieved first place in the FAME challenge with 19.9% EER.

Demonstrated robustness in multilingual environments.

Enhanced intra-modal and inter-modal feature learning.

Abstract

The human brain has the capability to associate the unknown person's voice and face by leveraging their general relationship, referred to as ``cross-modal speaker verification''. This task poses significant challenges due to the complex relationship between the modalities. In this paper, we propose a ``Multi-stage Face-voice Association Learning with Keynote Speaker Diarization''~(MFV-KSD) framework. MFV-KSD contains a keynote speaker diarization front-end to effectively address the noisy speech inputs issue. To balance and enhance the intra-modal feature learning and inter-modal correlation understanding, MFV-KSD utilizes a novel three-stage training strategy. Our experimental results demonstrated robust performance, achieving the first rank in the 2024 Face-voice Association in Multilingual Environments (FAME) challenge with an overall Equal Error Rate (EER) of 19.9%. Details can be found in https://github.com/TaoRuijie/MFV-KSD.