English

Multi-target DoA Estimation with an Audio-visual Fusion Mechanism

Sound 2021-05-14 v1 Robotics Audio and Speech Processing

Abstract

Most of the prior studies in the spatial \ac{DoA} domain focus on a single modality. However, humans use auditory and visual senses to detect the presence of sound sources. With this motivation, we propose to use neural networks with audio and visual signals for multi-speaker localization. The use of heterogeneous sensors can provide complementary information to overcome uni-modal challenges, such as noise, reverberation, illumination variations, and occlusions. We attempt to address these issues by introducing an adaptive weighting mechanism for audio-visual fusion. We also propose a novel video simulation method that generates visual features from noisy target 3D annotations that are synchronized with acoustic features. Experimental results confirm that audio-visual fusion consistently improves the performance of speaker DoA estimation, while the adaptive weighting mechanism shows clear benefits.

Keywords

Cite

@article{arxiv.2105.06107,
  title  = {Multi-target DoA Estimation with an Audio-visual Fusion Mechanism},
  author = {Xinyuan Qian and Maulik Madhavi and Zexu Pan and Jiadong Wang and Haizhou Li},
  journal= {arXiv preprint arXiv:2105.06107},
  year   = {2021}
}

Comments

ICASSP 2021 accepted

R2 v1 2026-06-24T02:04:02.004Z