Facial Expression Recognition using Hahn Moment on Facial Patches
Tobin T Kaleekal, Jyotsna Singh · 2019
Facial expression recognition involves the identification and extraction of facial features that have maximum discriminative ability between the different expression classes. It has been found that most of the information that discriminates between expressions is contained in very specific regions of the face. This paper attempts to extract facial features from very localized regions of the face image by first isolating specific regions and then evaluating localized Hahn moments at different points in these regions. A Point distribution model (PDM) is trained in the proposed method to get the best fit of the landmark pattern for any face image. These landmark points are then used to isolate various regions of the face (facial patches). Then, Hahn moments are evaluated in a targeted manner to specific locations of these facial patches. Hahn moments with controlled region of emphasis and degree of localization gives exceptionally good features for facial expression recognition. The final feature vector for a face image contains Hahn moment features from all the facial patches. These feature vectors are used to train support vector machine (SVM) classifiers that use a one-against-one (OAO) classification technique. These classifiers are used to test the effectiveness of the proposed method. Experiments were done on the extended Cohn-Kannade (CK+) dataset as well as JAFFE dataset.