Hololens 2 Hand Joints Dataset for Hand Gesture Recognition

The Hololens 2 Hand Joints Dataset for Hand Gesture Recognition was collected using a HoloLens 2 AR headset, which provides tracking data for 26 hand joints. Participants performed a set of predefined gestures designed to meet the requirements of the SUN project. The dataset includes four static gesture categories -Thumbs Up, Thumbs Down, Two Fingers, and No Gesture- as well as two dynamic gesture categories: Moving Two Fingers and Random.For static gestures, each frame contains 129 features: 25 values representing the distances from the root joint (palm) to each of the other joints, and 104 values representing joint rotations encoded as quaternions. Each frame is also timestamped. For dynamic gestures, the data consists of frame sequences spanning a duration of one second. Each frame in these sequences likewise contains 129 features, following the same representation as in the static gestures.