Facial Image Emotion Recognition Based on Convolutional Neural Networks and Haar Classifiers

Facial expression shows the richest human expression and mainly conveys emotions and social signals. In recent years, the development of artificial intelligence technology and sufficient data have broken through previous limitations, opening up the development of intelligent emotion recognition. In this study, emotion recognition is conducted by a deep learning model with multiple layers to describe global features of facial emotions with facial images as input data and neural networks to learn facial features such as eyebrows, eyes, and mouth. The proposed model objectively and quickly presents emotional results, making it applicable to customer service feedback, judgment basis for medical personnel, fatigue driving detection, and more. The model uses facial images as input into a Haar classifier to remove the background of the image and focus on capturing the facial region. Based on the Convolution Neural Network (CNN) and the FER-2013 (Facial Expression Recognition 2013) test dataset. After the user inputs the facial image, the system's prediction accuracy increased by 7.83% compared to the baseline system, effectively improving the accuracy of facial emotion recognition.