Preprint
Aug 2026
Identity-Faithful Audio-Visual Target Speaker Extraction with REAL-2MIX and VOXBLINK2-AVSE
REAL-2MIX, a Mandarin AV-TSE benchmark of jointly recorded real two-speaker mixtures with synchronized multi-view video, and REAL-2MIX-AVSE, a Mandarin AV-TSE benchmark of jointly recorded real two-speaker mixtures with synchronized multi-view video are introduced.
Pei-Jun Yang, Zhan Jin, Xiao-Yi Qin et al.
· 0 citations