A Survey on Physiological Signal-Based Emotion Recognition.

Area of Science:

• Psychology
• Computer Science
• Biomedical Engineering

Background:

• Physiological signals offer reliable, uncontrollable data for emotion recognition.
• Existing reviews often overlook crucial challenges specific to emotion recognition systems.
• A comprehensive understanding of these challenges is vital for developing robust emotion recognition models.

Purpose of the Study:

• To bridge the gap in existing literature by reviewing critical aspects of emotion recognition using physiological signals.
• To provide a detailed analysis of inter-subject data variance, data annotation, pre-processing, data splitting, and multimodal fusion techniques.
• To identify key challenges and future research directions in the field of physiological emotion recognition.

Main Methods:

• Review of existing literature on emotion recognition using physiological signals.
• Analysis of data annotation techniques and their comparative effectiveness.
• Examination of pre-processing methods tailored to specific physiological signals.
• Evaluation of data splitting strategies for enhanced model generalization.
• Comparison of various multimodal fusion techniques.

Main Results:

• Physiological signals are highly reliable for emotion recognition due to their involuntary nature.
• Inter-subject data variance significantly impacts model performance.
• Effective data annotation and pre-processing are crucial for accurate emotion recognition.
• Appropriate data splitting and multimodal fusion enhance model generalization and robustness.

Conclusions:

• Addressing inter-subject variance, annotation, pre-processing, splitting, and fusion is essential for robust emotion recognition.
• This review consolidates critical knowledge and highlights areas for future research.
• Further investigation into these specific challenges will advance the development of reliable emotion recognition systems.