While gesture recognition has been extensively studied, most existing approaches focus on classifying pre-segmented gestures rather than recognizing them in continuous streams. This limitation poses a challenge for real-world applications where gestures must be detected and classified in real-time with minimal errors and delays. Given this research gap, developing an additional benchmark that accounts for the specific challenges of continuous recognition was necessary. Furthermore, existing methods from the literature had to be adapted to achieve reliable results across key performance metrics essential for continuous gesture recognition. This talk addresses these challenges by proposing two novel approaches designed to improve the accuracy and robustness of gesture recognition in continuous settings.

In this introduction, it will be presented and discussed suitable approaches to build 3D natural interfaces for XR applications that work in continuous time and use reliable recognition methods (these approaches must commit a low amount of errors and have a real-time latency) of mid-air heterogeneous hand gestures as a fundamental input.

Dr. Marco Emporio is currently a postdoctoral researcher at the Department of Computer Science at the University of Verona, where he focusses on improving continuous hand gesture recognition systems based on hand skeleton tracking within the CollaborICE project. His main research interests center on human-computer interaction in eXtended Reality (XR) environments, 3D gesture recognition, and the development of natural interfaces for real-time applications. His most significant achievement during the PhD period, was creating novel approaches for building 3D natural interfaces that work in continuous time with reliable recognition methods, achieving low error rates and real-time latency for mid-air heterogeneous hand gestures.

During his doctoral studies, he spent six months at the University of Central Florida, where he collaborated with the ISUE group to develop a 3D gesture generator aimed at improving classification method scores and conducted experiments involving static and dynamic gesture datasets across four game rooms. He obtained my PhD in Computer Science in 2025 under the supervision of Prof. Andrea Giachetti, following his master's degree in computer science & engineering with a focus on Visual Computing, from the University of Verona.

His research journey began with developing novel solutions for single-handed deviceless object manipulation in immersive visualization environments [1], then evolved to focus on online recognition of heterogeneous gestures from 3D hand pose trajectories. He created a convolutional neural network that achieved better accuracy and false-positive ratios than state-of-the-art methods for gesture classification, which he successfully implemented in a real-time demo for HoloLens 2 [2], [3]. His work spans multiple domains including extended reality, machine learning, digital twins, and Industry 4.0 applications [4], [5]. He has been actively involved in organizing conferences like STAG 2024 and the SHREC '22 contest [6] and has published in journals including Computer Vision and Image Understanding, and IEEE Access.