Advancing Face-to-Face Emotion Communication: A Multimodal Dataset (AFFEC)

TL;DR

AFFEC addresses the challenge of authentic, face‑to‑face emotion understanding by introducing a multimodal, openly available dataset that captures EEG, eye tracking, GSR, facial signals, and Big Five personality data across 84 simulated dialogues and six emotions. Crucially, it differentiates Expressed $E_e$, Perceived $E_p$, and Felt $E_f$ emotions, enabling nuanced analysis of internal states and observer interpretations. Baseline analyses demonstrate discriminative signals across modalities, with arousal generally easier to predict than valence; multimodal fusion and personality information yield the strongest performance, highlighting the value of personalization in affective computing. The dataset’s BIDS structure and public availability, along with baseline pipelines, position AFFEC as a robust foundation for developing context‑aware, adaptive, and user‑specific emotion recognition in human–agent interaction and social robotics.

Abstract

Emotion recognition has the potential to play a pivotal role in enhancing human-computer interaction by enabling systems to accurately interpret and respond to human affect. Yet, capturing emotions in face-to-face contexts remains challenging due to subtle nonverbal cues, variations in personal traits, and the real-time dynamics of genuine interactions. Existing emotion recognition datasets often rely on limited modalities or controlled conditions, thereby missing the richness and variability found in real-world scenarios. In this work, we introduce Advancing Face-to-Face Emotion Communication (AFFEC), a multimodal dataset designed to address these gaps. AFFEC encompasses 84 simulated emotional dialogues across six distinct emotions, recorded from 73 participants over more than 5,000 trials and annotated with more than 20,000 labels. It integrates electroencephalography (EEG), eye-tracking, galvanic skin response (GSR), facial videos, and Big Five personality assessments. Crucially, AFFEC explicitly distinguishes between felt emotions (the participant's internal affect) and perceived emotions (the observer's interpretation of the stimulus). Baseline analyses spanning unimodal features and straightforward multimodal fusion demonstrate that even minimal processing yields classification performance significantly above chance, especially for arousal. Incorporating personality traits further improves predictions of felt emotions, highlighting the importance of individual differences. By bridging controlled experimentation with more realistic face-to-face stimuli, AFFEC offers a unique resource for researchers aiming to develop context-sensitive, adaptive, and personalized emotion recognition models.

Figures (7)

• Figure 1: An illustrative overview of the three types of emotions examined in this study: Expressed emotion ($E_e$), Perceived emotion ($E_p$), and Felt emotion ($E_f$).
• Figure 2: Example scenario presented before the video stimulus. Such contextual prompts prime emotional responses akin to natural interactions.
• Figure 3: Representative frame from a CREMA-D video stimulus. Dynamic facial expressions support ecological validity in emotion research.
• Figure 4: The 9-point scales used for describing the emotional arousal and valence scores.
• Figure 5: Participant in the experimental setup with EEG, GSR, and eye-tracking sensors, viewing stimuli on a monitor.