Society. Spaces. Screens

Abstract

In the era of smart cities, urban planning demands an empathetic understanding of human behavior to create environments that foster well-being and safety. This paper introduces a state-of-the-art facial emotion recognition (FER) system that leverages a hybrid Convolutional Neural Network-Temporal Convolutional Network (CNN-TCN) architecture to decode subtle emotional shifts from image sequences. Unlike traditional FER models that classify static emotions, our approach captures temporal dynamics, allowing for real-time assessment of public sentiment in urban spaces. By integrating advanced preprocessing techniques to mitigate challenges such as lighting variability and background interference, the system achieves robust performance across diverse datasets. This framework is applied to analyze pedestrian emotions in high-traffic public areas, identifying stress points and emotional trends that influence spatial usage. The results offer actionable insights for optimizing urban layouts, enhancing public safety, and designing spaces that intuitively respond to collective human emotions. This pioneering approach transforms urban planning from a static process into a dynamic, emotion-driven discipline, laying the groundwork for truly human-centric smart cities.