Active Speakers in Context

Current methods for active speaker detection focus on modeling audiovisual information from a single speaker. This strategy can be adequate for addressing single-speaker scenarios, but it prevents accurate detection when the task is to identify who of many candidate speakers are talking. This paper introduces the Active Speaker Context, a novel representation that models relationships between multiple speakers over long time horizons. Our new model learns pairwise and temporal relations from a structured ensemble of audiovisual observations. Our experiments show that a structured feature ensemble already beneï¬ ts active speaker detection performance. We also ï¬ nd that the proposed Active Speaker Context improves the state-of-the-art on the AVA-ActiveSpeaker dataset achieving an mAP of 87.1%. Moreover, ablation studies verify that this result is a direct consequence of our long-term multi-speaker analysis.

Alcazar, J. L., Caba, F., Mai, L., Perazzi, F., Lee, J.-Y., Arbelaez, P., & Ghanem, B. (2020). Active Speakers in Context. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). doi:10.1109/cvpr42600.2020.01248

Institute of Electrical and Electronics Engineers (IEEE)

Conference/Event Name
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)



Additional Links

Permanent link to this record