跳到正文
arXiv cs.AI· Zhuoyuan Tang, Tao Wang, Kai Yang, Li Yuan, Yuezhou Zhang, Rongrong Zhong, Yi Cai, Saskia C. Sanderson, Jaime Delgadillo, Richard Dobson, Angus Roberts·· 10 小时前AI 评分29

SIGMA:基于说话人分离与症状结构化的临床访谈抑郁严重程度多模态估计框架

SIGMA: speaker-isolated, symptom-structured multimodal estimation of depression severity from clinical interviews

AI 导读

SIGMA 通过说话人分离仅保留受访者语音,并用跨注意力让文本、音频、视频在编码时相互条件化,再以分级响应头联合预测八项问卷条目。在 Extended Distress Analysis Interview Corpus 上,SIGMA 取得 CCC 0.762、RMSE 4.13、MAE 3.30 的 SOTA 结果(五个随机种子)。

正文 · 原文

Authors:Zhuoyuan Tang, Tao Wang, Kai Yang, Li Yuan, Yuezhou Zhang, Rongrong Zhong, Yi Cai, Saskia C. Sanderson, Jaime Delgadillo, Richard Dobson, Angus Roberts

View PDF HTML (experimental)

Abstract:Clinical interviews are central to depression assessment but time-consuming and rater-dependent. Automated severity estimation offers scalable screening, yet existing methods often retain interviewer speech (confounding participant cues), fuse modalities late (losing fine-grained cross-modal dynamics), or score symptoms independently (ignoring inter-symptom correlations). We present SIGMA, a speaker-aware, symptom-structured framework for depression-severity estimation from semi-clinical interviews. SIGMA uses speaker diarization to restrict input to participant speech, cross-attention so that text, audio and video condition one another during encoding, and a graded-response head that predicts the eight questionnaire items jointly through a shared latent severity trait with item-specific ordered thresholds. On the Extended Distress Analysis Interview Corpus, SIGMA achieves state-of-the-art performance (CCC 0.762, RMSE 4.13, MAE 3.30 over five seeds). Ablations confirm that isolating participant speech and joint, structured symptom prediction each substantially improve concordance and estimation accuracy.
Subjects: Sound (cs.SD); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as: arXiv:2611.23348 [cs.SD]
  (or arXiv:2611.23348v1 [cs.SD] for this version)
  https://doi.org/10.48550/arXiv.2611.23348

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Tao Wang [view email]
[v1] Tue, 17 Nov 2026 11:56:15 UTC (1,414 KB)

来源:arXiv cs.AI · arxiv.org