EXPLORING CUSTOMER BEHAVIOR THROUGH EMOTION DETECTION

This thesis investigates the potential of emotion detection, gaze tracking, and hand movement analysis as tools to enhance understanding of customer behavior in retail environments. In retail, subtle indicators such as facial expressions, gaze direction, and hand gestures provide valuable insights into customer preferences and decision-making processes. The study leverages real-time video processing with a multithreaded approach on a Raspberry Pi-powered system embedded in a gift cart equipped with dual cameras. Using FER for emotion detection and a custom model for gaze and hand tracking, this research identifies customer engagement patterns, focusing on how emotions like happiness correlate with purchasing intent. The key research questions are structured to explore how each behavioral component influences overall customer experience. Specifically, gaze direction is analyzed to understand engagement, emotion recognition evaluates satisfaction, and hand movements indicate product interaction. The study also considers technical limitations, including processing speed constraints and low-light sensitivity, inherent to edge devices like the Raspberry Pi. These challenges are addressed with optimization strategies to maintain accuracy and real-time analysis. Results indicate that analyzing emotional responses and behavior patterns offers a robust, non-intrusive method for retailers to improve customer experience and tailor marketing strategies effectively. This thesis contributes to the intersection of computer vision, artificial intelligence, and retail analytics, aiming to provide a comprehensive framework for enhancing customer behavior analysis through emotion-driven insights.