Advances in Video Emotion Recognition: Challenges and Trends

Area of Science:

• Affective Computing
• Computer Vision
• Human-Computer Interaction

Background:

• Video emotion recognition (VER) analyzes viewer emotions from video content.
• Applications include video recommendation, HCI, and intelligent education.
• VER is grounded in psychological models of emotion.

Purpose of the Study:

• To provide a comprehensive review of VER.
• To analyze existing datasets, evaluation metrics, and algorithms.
• To identify current challenges and propose future research directions.

Main Methods:

• Analysis of psychological models foundational to VER.
• Review and categorization of VER algorithms.
• Comparison and analysis of classic methods across four datasets.
• Identification of challenges in emotional representation, datasets, and multimodal integration.

Main Results:

• Key challenges identified: emotional representation gaps, dataset limitations, and multimodal fusion.
• Classic VER methods were compared on standard datasets.
• The study highlights the need for advanced neural networks and fusion strategies.

Conclusions:

• Future research should focus on advanced neural networks and multimodal fusion.
• Developing high-quality emotional representations and active learning strategies is crucial.
• Addressing current challenges will advance the field of video emotion recognition.