A dataset of continuous affect annotations and physiological signals for emotion analysis

Area of Science:

• Affective computing and computational psychology.
• Physiological signal processing and machine learning for emotion recognition.

Background:

• Accurate emotion assessment is challenging due to the inability to conduct real-time, in-situ studies.
• Current methods often rely on indirect, post-hoc recordings, limiting ecological validity.
• The need for continuous, real-time emotion data collection in naturalistic settings is critical.

Purpose of the Study:

• To introduce the Continuously Annotated Signals of Emotion (CASE) dataset for real-time emotion analysis.
• To enable simultaneous annotation of emotional valence and arousal using an intuitive interface.
• To provide synchronized physiological recordings alongside continuous emotion annotations.

Main Methods:

• Development of a novel, joystick-based interface for real-time, continuous emotion annotation (valence and arousal).
• Acquisition of synchronized, high-frequency physiological data (ECG, BVP, EMG, GSR/EDA, respiration, skin temperature) from 30 participants.
• Utilized validated video stimuli to induce emotional responses in participants.

Main Results:

• The CASE dataset comprises synchronized physiological and continuous self-reported emotion data from 30 participants (15 male, 15 female).
• Demonstrated the validity of emotion induction through analysis of annotation and physiological data.
• The developed interface allowed for simultaneous, intuitive reporting of valence and arousal.

Conclusions:

• The CASE dataset offers a valuable resource for advancing real-time emotion recognition research.
• Continuous annotation and physiological data capture provide a more nuanced understanding of emotional dynamics.
• This dataset facilitates the development of more accurate computational models of emotion.