Biosignal-Based Multimodal Emotion Recognition in a Valence-Arousal Affective Framework Applied to Immersive Video...

Area of Science:

• Affective computing
• Human-computer interaction
• Physiological computing

Background:

• Emotion recognition is crucial for personalized user experiences.
• Existing methods vary in elicitation, data sources, and classification.
• Multimodal biosignal analysis offers a promising avenue for objective emotion assessment.

Purpose of the Study:

• To investigate the efficacy of multiple biosignals for emotion assessment during immersive video visualization.
• To develop a user-independent emotion recognition system.
• To evaluate classification performance for valence and arousal dimensions.

Main Methods:

• Collected multimodal data: Electrocardiography (ECG), Electrodermal Activity (EDA), Blood Volume Pulse (BVP), and Respiration.
• Extracted physiological and statistical features from biosignals.
• Employed Support Vector Machines (SVM) for user-independent classification into two classes per dimension (valence, arousal).
• Utilized Self-Assessment Manikin (SAM) for subjective emotional state reporting.

Main Results:

• Achieved 69.13% accuracy for arousal classification.
• Achieved 67.75% accuracy for valence classification.
• Results demonstrate the potential of multimodal biosignals for emotion recognition.

Conclusions:

• The proposed approach shows encouraging results for emotion assessment using biosignals.
• Further research with larger datasets and populations is recommended.
• This study contributes to advancing objective emotion recognition technologies.