Facial expression recognition based on local binary patterns and kernel discriminant isomap.

Area of Science:

• Computer Vision
• Machine Learning
• Pattern Recognition

Background:

• Facial expression recognition is a complex task due to the nonlinear structure of facial images.
• Existing dimensionality reduction techniques may not fully capture discriminant information for improved recognition.

Purpose of the Study:

• To propose Kernel Discriminant Isometric Mapping (KDIsomap), a novel kernel-based manifold learning method.
• To enhance nonlinear feature extraction for facial expression recognition by maximizing interclass scatter and minimizing intraclass scatter.

Main Methods:

• Kernel Discriminant Isometric Mapping (KDIsomap) was developed for nonlinear dimensionality reduction.
• Local Binary Patterns (LBP) were used for facial feature extraction.
• A nearest neighbor classifier with the Euclidean metric was employed for classification.

Main Results:

• KDIsomap achieved high accuracy rates: 81.59% on the JAFFE database and 94.88% on the Cohn-Kanade database.
• The method demonstrated superior performance compared to Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Kernel Principal Component Analysis (KPCA), Kernel Linear Discriminant Analysis (KLDA), and Kernel Isometric Mapping (KIsomap).

Conclusions:

• KDIsomap effectively extracts discriminant information from facial features, leading to significant improvements in facial expression recognition.
• The proposed method offers a robust approach for nonlinear dimensionality reduction in complex pattern recognition tasks.