Video-Based Facial Micro-Expression Analysis: A Survey of Datasets, Features and Algorithms

Area of Science:

• Psychology
• Computer Science
• Biomedical Engineering

Background:

• Micro-expressions are involuntary, transient facial expressions revealing concealed emotions, crucial for applications like lie detection.
• Their subtle nature and brief duration make detection and recognition challenging, often requiring expert experience.
• Video-based micro-expression analysis is an emerging research area due to its complexity and potential insights.

Purpose of the Study:

• To provide a comprehensive survey of video-based micro-expression analysis.
• To systematically review developments, datasets, algorithms, and applications in the field.
• To introduce a new dataset (MMEW) and perform unified comparisons of existing methods.

Main Methods:

• A cascaded structure approach guiding the survey, covering neuropsychological basis, datasets, features, spotting, recognition, applications, and evaluation.
• Analysis of basic techniques, advanced developments, and challenges in micro-expression research.
• Introduction and release of the Micro-and-Macro Expression Warehouse (MMEW) dataset.

Main Results:

• A systematic overview of state-of-the-art video-based micro-expression analysis techniques.
• Identification of limitations in existing micro-expression datasets and introduction of the MMEW dataset.
• Unified comparison of representative spotting and recognition methods on established and new datasets.

Conclusions:

• The survey provides a unified evaluation and systematic overview of micro-expression analysis.
• The new MMEW dataset offers expanded resources for micro-expression research.
• Future research directions are outlined to advance the field of micro-expression analysis.