Modelling an automatic facial emotional valence detection system using deep convolutional neural network

Abstract

Human Emotion Recognition (HER) system with a computerized approach is considered as a powerful that assists in solving the complex problems in a wider range of real-time applications like healthcare, marketing, education and working environment. Various existing approaches make use of live video or digital images to trace the facial expressions of an individual from the group and intend to predict the emotional state of that person. The research work explained in this thesis explores the combinations of various facial emotion predicting approaches used by the researchers and the gaps identified during the prediction process is filled using the modern and an advanced approach known as Deep Learning (DL). The DL-based classifier model paves the way to handle the short comings identified in the existing image processing and Machine Learning (ML) approaches. The significance advantage of adopting deep learning approach is its tendency to handle the feature extraction and classification in an efficient manner. Thus, it reduces the computational complexity and provides better prediction accuracy. newlineFacial Emotion Recognition (FER) is a modern research area that deals with the classification of human emotions based on their facial expressions. Their facial expressions are used in various applications like intelligent human-computer interaction, biometric security, clinical medicine for mental health problem, pain, depression and autism, and robotics. This dissertation investigates the emergent deep learning techniques for analyzing the facial expression and designs artificial intelligent systems for practical and real-time applications. newline