Deep Learning based Facial Expression Recognition and its Applications

Abstract

It is more crucial than ever to have the capability to recognize emotions in today s newlineworld, which is dominated by artificial intelligence (AI) and the Internet of Things newline(IoT). Human-computer interaction, robotics, healthcare, biometric security, and behavioral modeling all stand to gain greatly from this. Facial expressions, text data, newlinebody movements, speech, and brain or heart impulses are all sources that can be used newlineby emotion detection systems to identify the emotions being expressed. Attitude, emotional newlinecontrol, and the ability to activate emotion are all relevant factors to consider newlinewhile studying feelings, along with the basic emotions. However, the development and newlineuse of Deep Learning algorithms have greatly enhanced recognition accuracy in an unconstrained environment. newlineEven though several techniques have been looked into, there are still problems, newlinesuch as changes in lighting, rotations, and occlusions. This research began by investigating whether it would be possible to use transfer learning to optimize the parameters of models that had not been pre-trained for the task of image classification. These models included the Visual Geometric Group (VGG) model and convolutional neural newlinenetworks (CNNs). Two channels of expression-related variables are extracted by the newlineVGG-19 network and the convolutional network from the grayscale face images using newlinethese networks. The linear fusion approach takes a weighted average of the classification choices that were made on training data from both channels and uses it to arrive at a conclusion about which class something belongs to. The ultimate recognition is achieved by using a convolutional neural network architecture in conjunction with a softmax classifier. There are seven different facial expressions of emotion (BEs), which are as follows:happiness, surprise, anger, sadness, fear, disgust, and neutral. newlineOn the NVIDIA Jetson Nano, a different method for facial emotion recognition with a high video handling outline rate employs the xception model in convolutional neura