EmoWear: Wearable Physiological and Motion Dataset for Emotion Recognition and Context Awareness

Area of Science:

• Biomedical Engineering
• Affective Computing
• Human-Computer Interaction

Background:

• Emotion Recognition (ER) research often relies on facial expressions or self-reports.
• Seismocardiography (SCG), measuring chest wall vibrations via Inertial Measurement Units (IMUs), offers a less intrusive method for physiological monitoring.
• Integrating SCG with other biosignals can enhance the accuracy and scope of ER.

Purpose of the Study:

• Introduce the EmoWear dataset for exploring Emotion Recognition (ER) using Seismocardiography (SCG).
• Facilitate multi-modal ER and context-aware sensing (gait, voice, drinking) using a single IMU.
• Provide a validated dataset for advancing ER research through physiological signals.

Main Methods:

• Collected multi-modal physiological data (SCG via IMUs, ECG, BVP, EDA, SKT) from 49 participants during emotionally stimulating video stimuli.
• Recorded participant self-assessments of emotional valence, arousal, and dominance.
• Acquired data during activities including walking, talking, and drinking for contextual awareness.

Main Results:

• Demonstrated the effectiveness of emotional stimulation through statistical analysis.
• Verified the quality of collected physiological signals using signal-to-noise ratio and correlation analyses.
• Established the potential for ER via SCG and multi-modal ER using the EmoWear dataset.

Conclusions:

• The EmoWear dataset is a valuable resource for ER research, particularly using SCG.
• The dataset supports the development of context-aware systems leveraging IMUs for sensing emotions, gait, voice, and drinking.
• EmoWear advances the field of physiological-based emotion recognition and multi-modal sensing.