Real-Time Facial Emotion Recognition Using Haar-Cascade Classifier and MobileNet in Smart Cities

Smart city applications aim to improve the quality of life for citizens by using digital technologies. Facial emotion recognition (FER) has become essential for many smart city applications as the facial emotions of the citizens are subjective data that can be properly managed to enhance many services in a smart city. FER helps in recognizing the state of human emotions such as happiness, surprise, neutrality, sadness, disgust, fear, and anger based on facial photos or videos. The main challenge in FER is to automatically distinguish different facial emotion states with precise accuracy in real-time. In this paper, the proposed approach identifies citizens’ facial emotions by combining the OpenCV Haar-Cascade classifier for automatic real-time face detection and the MobileNet as an emotion detection model. This research is concerned with the sustainable development goals (SDGs) related to cities and communities. Specifically, SDG 11 “Sustainable Cities and Communities”, which aims to provide a high-level quality of life for all citizens. The accuracy of the proposed approach has been tested on the popular FER2013 dataset of facial expressions, and it achieves promising accuracy for the given dataset and is proven to be applicable in real-world scenarios.