Distributed Edge Computing for Appearance-Based Gait Recognition

The expansion of smart cities requires the integration of advanced security technologies, including gait recognition, which is increasingly valued for enhancing urban surveillance. This paper introduces a novel distributed edge computing framework specifically designed for appearance-based real-time gait recognition, utilizing deep learning techniques to ensure efficient and accurate analysis. The framework leverages a distributed architecture that minimizes latency and distributes computational loads across multiple devices, effectively balancing processing demands. We also introduce the OAKGait16 dataset, which comprises video sequences of various individuals from multiple view angles and in different clothing variants, providing a robust basis for comprehensive analysis. Our evaluations demonstrate that our approach not only achieves high efficiency and accuracy in controlled settings-but also keeps the processing speed at 29 frames per second, showcasing its potential for real-world application in smart city security systems. This work establishes a foundational framework intended to guide further advancements in practical, scalable gait recognition systems.