DYLEM-HGR - A Dataset for Dynamic Hand Gesture Recognition with Leap Motion

Dynamic gesture recognition, which involves the interpretation of fluid- and time-dependent movement patterns, is a cornerstone of human-computer interaction (HCI). Unlike static gestures, dynamic gestures require systems capable of tracking and processing continuous motion in real time, making their recognition challenging. This is particularly critical in virtual reality (VR) and healthcare applications, where intuitive, touchless interfaces can significantly enhance user experience and operational efficiency. The effectiveness of gesture recognition systems is based on the availability of significant and reliable gesture datasets. In this paper, we present a dataset for Dynamic Hand Gesture Recognition with Leap Motion (DYLEM-HGR), featuring 400 dynamic hand gestures recorded using a Leap Motion Controller, providing rich, detailed data on hand and finger movements. The DYLEM-HGR is made available in three versions: i) as raw time series; ii) as cleaned and scaled time series; and iii) as statistic, engineered features. This allows the scientific community to facilitate the development of more sophisticated and accurate machine learning algorithms for gesture recognition. The DYLEM-HGR aims to catalyse innovation in dynamic gesture recognition, fostering more intuitive and effective HCI systems.