Multi-Directional Ternary Pattern for Facial Expression Recognition
Kiran Sudam Dorage, Smitha Raveendran · 2019
Multi-Directional Ternary Pattern (M-DTP) is the face descriptor which reduces the noise and extracts the key expression related characteristics of the input face image to overcome the recent challenges encountered in unconstrained face recognition. In M-DTP, the derivative of the Gaussian operator is utilized to transform the gray scale facial image into the multi-directional gradient image that is vigorous to changes in the illumination. This pre-processing is introduced to minimize the challenges due to the variations in illumination and unconstrained face expression recognition. Further, by making use of directional information and ternary pattern the emotion-related features are extracted from the face. The proposed multi-level histogram features extracted from the ternary face code technique delivers the finer grain description of facial emotions to improve the facial expression performance under various illumination conditions. The features extracted using M-DTP are high-dimensional, therefore to reduce the dimensionality Principal Component Analysis (PCA) is used. The Support Vector Machine (SVM) classifier is utilized to perform and evaluate the classification performance using different datasets.