Low-Cost Video--Time Priors as a Strong Baseline for EEG--fNIRS Emotion Regression on Familiar Videos

2026-10-02

AI Summary

This study addresses the limited contribution of physiological signals in EEG-fNIRS-based emotion regression within familiar video contexts. We propose leveraging video temporal priors as a strong baseline model, repositioning physiological signals as optional residual correction terms. The approach is systematically evaluated using leave-subject-out five-fold cross-validation, source-explicit ablation studies, and fixed fusion strategies. Results demonstrate that the low-cost video prior outperforms complex multimodal fusion schemes, achieving mean absolute errors (MAE) of 0.05 and 0.32 for intra- and inter-subject evaluations, respectively. These findings confirm that video identity and temporal information are core factors in continuous emotion prediction, providing an efficient paradigm reference for multimodal affective computing.

Abstract

Continuous emotion regression estimates moment-to-moment valence and arousal while a viewer watches a video. In familiar-video deployment, responses fron training participant-specific estimate, and prior-dominating fixed fusion tests whether physiology adds residual correction. In five-fold subject-held-out evaluation on 24was within 0.05 and 0.32 MAE of fusion in the internal and external evaluations, respectively. Source-explicit ablations showed that video identity and within-video tine accounted for most of the reduction, while EG-FNIRS gains were smaller and varied across participants and videos. These results identify the video-time prior as a strong, low-cost baseline and position EEG-fNIRS as an optional residual signal for familiar-video emotion regression.

Problem

• continuous emotion regression
• EEG-fNIRS
• valence and arousal
• video-time priors
• familiar videos

Innovation

• Video-Time Priors
• EEG-fNIRS
• Emotion Regression
• Residual Correction
• Source-Explicit Ablations