Robust Facial Expression Recognition using Gabor and LDP Feature Fusion using CCA
Mahesh M Goyani · INTERNATIONAL JOURNAL OF NEXT-GENERATION COMPUTING · 2019
Facial Expression recognition has become vital considering its numerous applications including Human Computer Interaction, security, medical field etc. In order to effectively implement these applications, the proposed method aims to increase the overall accuracy and robustness of the recognition system. Fusion of two feature extraction methods, namely Gabor and Local Directional Pattern (LDP) that are complementary in nature is carried out. Gabor Features focuses on the structural details whereas LDP targets the textural and subtler details in a robust manner. Fusion is carried out using Canonical Correlation Analysis (CCA) based fusion, as it effectively utilizes the correlation between the two features for fusion. For the efficient training of the classifiers, optimal feature vector is obtained by projecting the original feature vector on Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) subspace. To prove the robustness of the proposed method, it is tested on benchmark datasets under ideal as well as different conditions such as noisy environment, low resolution, small sample space, different facial components etc. The method has shown better performance than the state of the art methods.