Gesture recognition benchmarking for operator support in Industry 5.0

With the advent of Industry 5.0, the convergence of human intelligence and advanced automation is redefining industrial operations. Gesture recognition technology has emerged as a pivotal tool in this paradigm, facilitating seamless collaboration between human operators and intelligent systems. In this sense, the growing role of gesture recognition in Industry 5.0 emphasizes its potential to enhance productivity, safety, and efficiency on the factory floor. In this context, different artificial intelligence solutions have been adopted to perform gesture recognition. This thesis aims to compare different deep learning-based approaches to recognize operator’s posture (Human Pose Estimation) and track operator’s movements in view of preserving and guaranteeing workers’ well-being. In particular, the proposal is to evaluate the accuracy and real-time performance of skeleton-based algorithms in order to comprehend the actual necessity to extract the temporal and/or visual features in the input data. In an era where the global workforce is continuously evolving, the success and sustainability of organizations are deeply intertwined with the physical, mental, and emotional health of their employees. For this reason, this thesis delves into innovative digital solutions adopting a user-centered approach to provide assistance to operators during their working activity. This work includes an introduction on the context of the thesis, theoretical notes on deep learning, a description of the proposed method and the results obtained.