An Optimal MSER Descriptor Based Facial Expression Recognition System using Artificial Intelligence Method

Abstract

In this Artificial intelligence based Facial emotion recognition system (AI_FERS) model, emotions of facial expressions through performing some predefined steps such as face acquisition, pre-processing of images, face detection, feature extraction & classification have recognized. In the pre-processing of the image phase include the approaches used for face detection is: Knowledge-based, Feature-based, Template-based, and Appearance-based approach. Binary image computation, Skin-color segmentation and morphological filtering, which includes the dilation of Binary images and Gray Images are being extensively applied. For features extraction from images MSER (Maximally Stable External Regions) technique is used. At the final step categorize of emotion into six parts: surprise, fear, disgust, anger, happiness, and sadness come as an outcome using ANN (Artificial Neural Network) technique. The efficiency of the system is examined based on performance parameters such as FAR, FRR, accuracy and execution time. The average accuracy of the AI_FERS model examined is about 98.23 %.