MUPA²S multimodal framework assesses stress with facial video and ECG
Original title: MUPA${}^2$S: Multimodal Unified Perception with Asymmetric Attention for Stress Assessment
MUPA²S unified Video-ECG fusion achieves 69.58% accuracy on 58-subject dataset at half the parameters of feature-level fusion.
MUPA²S combines facial video and ECG signals in a unified perception architecture without modality-specific components. Asymmetric attention comprises segment-local cross-attention and global self-attention across segment states. Video-only achieves 69.44% accuracy while ECG-only remains near chance. Unified fusion provides parameter-efficient stress assessment.
- 58-subject stress dataset collected
- Unified Video-ECG fusion at 69.58% accuracy
- Half the parameters of feature-level fusion
- Video modality outperforms ECG modality
From the original article
1 min read at arxiv.orgStress is a leading contributor to long-term physical and mental health decline. The automatic assessment of stress is a promising approach for continuous monitoring and mitigating human biases. However, the development of automated systems remains challenging due to the heterogeneity of physiological and behavioral signals.
By Stefanos Gkikas · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.