Identity-Faithful Audio-Visual Target Speaker Extraction with QIANGDA and VOXBLINK2-AVSE

Abstract

Audio-visual target speaker extraction should return the speaker indicated by the video, yet a separator can ignore the visual cue and repeatedly output the acoustically dominant voice. We introduce QIANGDA, a Mandarin AV-TSE benchmark of jointly recorded real two-speaker mixtures with synchronized multi-view video. Each scene also contains preceding A-only and B-only stages that provide in-scene speaker references. It contains 77 scenes and 7,598 clips (11.84 hours), including 6,042 dual-annotated mixtures. After processing there leave 6,038 evaluable mixtures and 12,076 target-speaker rows. We additionally curate VOXBLINK2-AVSE from VoxBlink2, comprising 250,828 synchronized audio--lip-ROI pairs from 28,421 identities and 766.17 hours of speech. Our extractor uses frozen, 1,280-dimensional projected AV-HuBERT features, target-conditioned training, and layer-wise feature modulation. We jointly evaluate content with Qwen3-ASR-1.7B CER and target identity with WeSpeaker ResNet34 plus Overlapped Speech Detection (OSD). On the complete manifest, the best archived checkpoint obtains 0.2261 CER, 82.22% strict output correctness, and 69.53% both-output strict success.