Convolutional Neural Network based Human Emotion Recognition System: A Deep Learning Approach

Recent research focuses towards Expression recognition. Variety of applications is now available ranging from security cameras to detecting emotions. Facial recognition is an important activity in emotion detection Convolutional Neural Networks (CNN) are used for facial recognition. Images are taken as input and facial expressions are produced as outcome like Happy, Sad, Disgust, Angry, Fear, Surprise and neutral. In this paper, we propose an Artificial Intelligence (AI) which recognizes the facial emotions using the different layers in the CNN. Thorough examination of deep Face Expression Recognizer (FER), including datasets and methods that shed light on these underlying difficulties. First, the FER scheme, which includes pertinent background information, is implemented for seeking advice for each level. The dataset used for experimentation is FER challenge dataset available in kaggle repository. The implementation environment includes keras, tensorflow, cv2 python packages. The results include the comparison of accuracy of emotion detection between training and testing phase. The average accuracy achieved was 84.50%.