A Wearable Motion Capture Dataset for Gait Analysis Using IMUs and Shank-Mounted Egocentric Cameras

Walking kinematics offer critical insights into human gait, enabling clinicians to assess motor function, diagnose impairments, and monitor rehabilitation progress. With the rise of machine learning, wearable sensors are increasingly used to estimate joint kinematics. To advance these fields, we introduce a dataset comprising 14 walking trials performed by 10 healthy participants across various speeds and trial types on overground, treadmill, slope, and stair locomotion modes. The dataset includes inertial measurement unit (IMU) data and shank-mounted egocentric video, synchronized with joint kinematics derived from optical motion capture and musculoskeletal modeling. It provides 327 minutes of recordings, including IMU signals, motion capture marker trajectories, joint angles, and 588k egocentric video frames per foot with corresponding histogram of optical flow (HOF) features. To the best of our knowledge, this is the first dataset offering multimodal wearable IMU and shank-mounted video data with ground-truth joint kinematics across a wide range of walking tasks. This dataset is intended to support a wide range of applications in both biomechanics and machine learning research.