Sparse Spatiotemporal Descriptor for Micro-Expression Recognition Using Enhanced Local Cube Binary Pattern.

Area of Science:

• Computer Science
• Psychology
• Biometrics

Background:

• Micro-expressions are spontaneous facial expressions revealing psychological states.
• Recognizing micro-expressions is crucial for clinical diagnosis, psychological research, and security.
• Challenges include the short duration and low intensity of micro-expressions.

Purpose of the Study:

• To develop a robust descriptor for micro-expression recognition.
• To address the challenges posed by subtle and fleeting facial movements.
• To enhance the accuracy of automated micro-expression detection systems.

Main Methods:

• A novel Enhanced Local Cube Binary Pattern (Enhanced LCBP) descriptor was developed.
• Enhanced LCBP integrates three complementary features: Spatial Difference LCBP, Temporal Direction LCBP, and Temporal Gradient LCBP.
• Multi-Regional Joint Sparse Learning was used for feature selection, focusing on critical regions.
• Multi-kernel Support Vector Machine (SVM) was employed for feature fusion and classification.

Main Results:

• The proposed Enhanced LCBP descriptor effectively captures spatiotemporal complementarity for subtle facial changes.
• Multi-Regional Joint Sparse Learning successfully reduced redundant information, improving discriminative ability.
• The method achieved promising results and demonstrated significant advantages on four spontaneous micro-expression datasets.
• Parameter evaluation and confusion matrix analysis confirmed the method's sufficiency and effectiveness.

Conclusions:

• The developed Enhanced LCBP descriptor offers a powerful tool for micro-expression recognition.
• The combination of Enhanced LCBP, sparse learning, and multi-kernel SVM provides a highly effective approach.
• This research advances the field of micro-expression analysis with practical implications for various applications.