GAITEX: Human motion dataset of impaired gait and rehabilitation exercises using inertial and optical sensors

TL;DR

This work presents a multimodal dataset of physiotherapeutic and gait-related exercises, including correct and clinically relevant variants, recorded from 19 healthy subjects using synchronized IMUs and optical marker-based motion capture (MoCap).

Abstract

Wearable inertial measurement units (IMUs) provide a cost-effective approach to assessing human movement in clinical and everyday environments. However, developing the associated classification models for robust assessment of physiotherapeutic exercise and gait analysis requires large, diverse datasets that are costly and time-consuming to collect. We present a multimodal dataset of physiotherapeutic and gait-related exercises, including correct and clinically relevant variants, recorded from 19 healthy subjects using synchronized IMUs and optical marker-based motion capture (MoCap). It contains data from nine IMUs and 68 markers tracking full-body kinematics. Four markers per IMU allow direct comparison between IMU- and MoCap-derived orientations. We additionally provide processed IMU orientations aligned to common segment coordinate systems, subject-specific OpenSim models, inverse kinematics outputs, and visualization tools for IMU-derived orientations. The dataset is fully annotated with movement quality ratings and timestamped segmentations. It supports various machine learning tasks such as exercise evaluation, gait classification, temporal segmentation, and biomechanical parameter estimation. Code for postprocessing, alignment, inverse kinematics, and technical validation is provided to promote reproducibility.

Figures (7)

• Figure 1: Top view of the experimental setup including treadmill, safety equipment, resistance band frame, and camera configuration. Eight Qualisys cameras were used for marker tracking, and two RGB cameras for video recording.
• Figure 2: Anterior (left) and posterior (right) views of the experimental marker setup. IMUs are shown in blue, Qualisys markers used for segment pose tracking are marked in red, and Qualisys markers mounted on IMUs for optical orientation validation are shown in orange. Marker and sensor labels match the naming conventions used in the dataset. Note that the IMU placed on the right forefoot is labeled as XSens_Hand_Right, since the standard Xsens setup does not include a dedicated IMU for this position. Therefore, the otherwise unused right-hand IMU was repurposed for the right forefoot.
• Figure 3: Overview of the physiotherapeutic exercises and their execution variants. Both exercises are performed in a standing position. The first exercise RD is illustrated in the upper row, and the second exercise RGS is illustrated in the lower row. For each exercise, the first image from the left illustrates the standardized start position. The second image depicts the correct execution according to physiotherapeutic guidelines. The remaining three images on the right display common erroneous movement patterns observed in individuals with foot drop. Detailed descriptions of the respective deviations can be found in Table \ref{['tab:exercise_protocol']}.
• Figure 4: Illustration of the two gait-related tasks. The left image shows a participant walking on a treadmill during the NG condition, without any assistive device. The right image depicts the same participant performing the GWO task under otherwise identical conditions, while wearing a rigid knee orthosis on the right leg that restricts knee flexion to 0°.
• Figure 5: Schematic illustration of the transformation chain used to align IMU orientations with their corresponding OpenSim body segments. The transformation $^{\mathrm{XKF3hm}}_{\mathrm{Segment}}T$ is decomposed into three components: a fixed rotation $^{\mathrm{XKF3hm}}_{\mathrm{OpenSim_y}}T$ aligning the vertical axes of the XKF3hm and OpenSim CS; a heading correction $^{\mathrm{OpenSim_y}}_{\mathrm{OpenSim}}T$ accounting for yaw differences between initial IMU and marker-based orientations; and the segment-specific transformation $^{\mathrm{OpenSim}}_{\mathrm{Segment}}T$, derived from the initial pose calibration. All transformations are represented as quaternions.