Efficient spatio-temporal local binary patterns for spontaneous facial micro-expression recognition.

Area of Science:

• Computer Science
• Artificial Intelligence
• Biomedical Engineering

Background:

• Micro-expression recognition is crucial for clinical diagnosis and deceit analysis but faces dataset development challenges.
• Existing methods like LBP-TOP, while effective, lack compactness for efficient feature extraction.
• Publicly available datasets such as SMIC and CASME II are vital for spontaneous micro-expression research.

Purpose of the Study:

• To propose two novel and efficient feature extraction approaches for micro-expression recognition.
• To improve the compactness and reduce redundancy in feature encoding compared to LBP-TOP.
• To enhance the accuracy and efficiency of spontaneous micro-expression recognition.

Main Methods:

• Development of LBP-Six Intersection Points (SIP) for feature extraction.
• Introduction of LBP-Three Mean Orthogonal Planes (MOP) as a super-compact feature descriptor.
• Comprehensive experimental evaluation of the proposed methods on spontaneous micro-expression datasets.

Main Results:

• The proposed LBP-SIP and LBP-MOP methods demonstrate improved recognition accuracy.
• Both approaches offer significant reductions in feature redundancy and improved compactness.
• Experimental results validate the efficiency and effectiveness of the novel feature extraction techniques.

Conclusions:

• The LBP-SIP and LBP-MOP methods provide a more compact and discriminative feature representation for micro-expression recognition.
• These novel approaches address the limitations of existing methods, paving the way for more efficient and accurate micro-expression analysis.
• The findings contribute to advancing the field of micro-expression recognition, particularly for spontaneous expressions.