Mood as a Contextual Cue for Improved Emotion Inference
CoRR(2024)
摘要
Psychological studies observe that emotions are rarely expressed in isolation
and are typically influenced by the surrounding context. While recent studies
effectively harness uni- and multimodal cues for emotion inference, hardly any
study has considered the effect of long-term affect, or mood, on
short-term emotion inference. This study (a) proposes time-continuous
valence prediction from videos, fusing multimodal cues including
mood and emotion-change (Δ) labels, (b) serially
integrates spatial and channel attention for improved inference, and (c)
demonstrates algorithmic generalisability with experiments on the EMMA
and AffWild2 datasets. Empirical results affirm that utilising mood
labels is highly beneficial for dynamic valence prediction. Comparing
unimodal (training only with mood labels) vs multimodal (training
with mood and Δ labels) results, inference performance improves for the
latter, conveying that both long and short-term contextual cues are critical
for time-continuous emotion inference.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要