Dynamic Background Inertial Gesture Dataset

 This dataset contains synchronized accelerometer and gyroscope data  from 5 participants performing 0-9 hand gestures during standing, walking and running conditions. Each sample includes precise temporal annotations in seg_label.txt indicating gesture onset/offset times (in seconds), specifically designed for training CAM-based attention models to localize gestures in dynamic environments. The dataset provides raw inertial signals with movement metadata while preserving participant anonymity.