Facial emotion recognition using Handcrafted features and CNN

Abstract

Emotion recognition technology is a form of face detection and recognition that takes an aid of facial expressions and biophysical signs and symptoms like pulse rate and activations in the brain to actually know about an individual state. Image-based facial expression identification is a challenging topic, especially when it comes to assessing human emotion or mood in certain situations, such as while enjoying or watching a series, or movie, getting engrossed in video games, while shopping, and even on the battlefield. Emotions are of utmost concern due to an immediate increase in a number of healthcare concerns such as depression, cancers, paralysis, trauma etc. This research proposes an approach to emotion detection using feature extraction and convolution neural networks. Feature extraction techniques used are HOG and SIFT coupled with the convolutional neural network. Datasets considered for the training and testing of the proposed approach are CKplus and JAFFE . Experimental results show that HOG with CNN surpasses the state-of-the-art models.