A Micro-GA Embedded PSO Feature Selection Approach to Intelligent Facial Emotion Recognition

Area of Science:

• Computer Vision
• Machine Learning
• Artificial Intelligence

Background:

• Facial expression recognition (FER) is crucial for human-computer interaction.
• Conventional methods often struggle with feature optimization and premature convergence.
• Developing robust and accurate FER systems remains a significant challenge.

Purpose of the Study:

• To propose an enhanced facial expression recognition system.
• To introduce a novel optimization algorithm, mGA-embedded PSO, for improved feature selection.
• To address the premature convergence issue prevalent in standard PSO algorithms.

Main Methods:

• Utilized modified local binary patterns for initial facial representation.
• Developed a micro genetic algorithm-embedded particle swarm optimization (mGA-PSO) for feature optimization.
• Incorporated advanced techniques like nonreplaceable memory and a secondary swarm within mGA-PSO.
• Employed multiple classifiers for recognizing seven distinct facial expressions.

Main Results:

• The proposed mGA-embedded PSO significantly mitigates premature convergence.
• Empirical results demonstrate superior performance compared to conventional PSO, GA, and other state-of-the-art FER models.
• Achieved high accuracy in facial expression recognition across within- and cross-domain datasets.

Conclusions:

• The developed system offers a significant advancement in facial expression recognition.
• The mGA-embedded PSO is an effective approach for feature optimization in complex pattern recognition tasks.
• The proposed method shows strong potential for real-world applications requiring accurate emotion detection.