Facial Expression Recognition Using Geometric Landmark Points and Convolutional Neural Networks

Facial Expression Recognition (FER) is a prevalent research area in human-computer interaction. In general, any FER technique contains three steps: feature extraction, dimensionality reduction, and classification. Dimensionality and feature selection are major issues concerning FER. Huge amount of memory and processing are required to process image as a whole. A more suitable alternative to this problem is geometric features. In this paper, facial landmark detection has used for feature extraction and Convolutional Neural Network Classifier. The method has tested on publicly available benchmark datasets: JAFFE, MUG, CK, and MMI. The results with the proposed methodology outperform the existing methods found in the literature.