Files
icassp2027-experiment/literature_notes/39_gaze_triadic_turn_taking.md
2026-08-30 15:36:28 +08:00

14 lines
980 B
Markdown
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# Gaze-Enhanced Multimodal Turn-Taking Prediction in Triadic Conversations
- 作者:Seongsil Heo、Christi Miller、Calvin Murdock、Michael Proulx
- 发表:Interspeech 2025
- 阅读深度:重点输入、ConvLSTM/user attention、评测章节
- 记录日期:2026-08-28
- 来源:[ISCA](https://www.isca-archive.org/interspeech_2025/heo25_interspeech.html)、[论文](https://www.isca-archive.org/interspeech_2025/heo25_interspeech.pdf)
## 摘要与项目关系
以 gaze、发声位置和 VAD 构建单人表示,再通过跨用户注意力建模三人交互。提供视觉方向与说话角色相关的证据,但使用 Aria 眼动、多通道音频及头运动,而非我们的视频混合单声道。
其 transition-level 评测包含转变后 1s,不能据此宣称已经实现提前 200ms 开口预测。我们可以借鉴逐人时序结构;不能把视频头朝向估计等同于眼动仪真值,也不能把角色识别指标与未来 onset AP 混用。