Design and Implementation of a Facial Expression Recognition and Emotion Analysis System Based on Deep Learning

Xinyu Xia, Bangying Wu, Kan Ji

Abstract: In response to the growing mental health concerns faced by individuals under high-pressure environments in modern society, as well as the limitations of effectively monitoring emotion states through traditional methods, this paper leverages deep learning techniques based on a modified VGG model to design and implement a facial expression recognition and emotion analysis system that can be transplanted to mobile platforms. This system has four functional modules including login & registration, emotion center, facial expression recognition & conversion, and personal center. In our study, the publicly available dataset FER-2013 was used to train the proposed model for recognizing and classifying various user emotions, covering a total of 7 emotional categories: anger, surprise, disgust, fear, happy, neutral, and sadness. In the system, facial expression data is collected through two methods: real-time photo capture and image import from album. Based on the acquired images, the system performs facial expression recognition and provides feedback on the detected emotional state to users. The results indicate that the proposed system not only provides users with convenient, real-time emotion detection via mobile devices, but also maintains good performance under low computational resource consumption.

Keywords: Mental Health, Deep learning, VGG, Facial expression recognition, Emotion analysis