University of Idaho - Physical Rehabilitation Movements Data Set (UI-PRMD)

A. Vakanski 1, D. Paul 2, R. Baker 2, H-p. Jun 2

1 University of Idaho, Industrial Technology, Idaho Falls, USA

2 University of Idaho, Department of Movement Sciences, Moscow, USA



Welcome to the UI-PRMD Web Site

UI-PRMD is a data set of movements related to common exercises performed by patients in physical therapy and rehabilitation programs. The data set is released under the Open Data Commons Public Domain Dedication and License (PDDL) v1.0, and as such is publically available for free to all interested users.

The movement data was collected in the Integrated Sports Medicine Movement Analysis Laboratory (ISMMAL) with the Department of Movement Sciences at the University of Idaho

The data set consists of 10 rehabilitation movements. A sample of 10 healthy individuals repeated each movement 10 times in front of two sensory systems for motion capturing: a Vicon optical tracker, and a Kinect camera. The data is presented as positions and angles of the body joints in the skeletal models provided by the Vicon and Kinect mocap systems.


The following 10 movements were selected for the data set: (1) deep squat, (2) hurdle step, (3) inline lunge, (4) side lunge, (5) sit to stand, (6) standing active straight leg raise, (7) standing shoulder abduction, (8) standing shoulder extension, (9) standing shoulder internal-external rotation, and (10) standing shoulder scaption.











Data Description

The data is organized into two folders ‘Vicon’ and ‘Kinect’, each containing the measurements acquired by either of the two respective sensory systems. Each folder contains two subfolders, ‘Positions’ and ‘Angles,’ which include the files with the respective values for the joint displacements.

The nomenclature of the files in the data set includes the following information: movement number_subject number_positions/angles. For example, the data instance ‘m04_s06_angles’ pertains to the 4th movement (i.e., side lunge) performed by the 6th subject, and it consists of the angular displacements of the joints expressed in degrees. Similarly, ‘m08_s02_positions’ corresponds to the recorded Cartesian position coordinates for the 2nd subject while performing the standing shoulder extension movement expressed in millimeters.

Further, because each movement consists of 10 episodes, or repetitions, the data is also provided in a segmented form, where each file comprises the measurements for one episode of one of the movements. The corresponding data is provided in the 'Segmented Movements' folder. The following file nomenclature is used for the segmented movements: movement number_subject number_episode number_positions/angles. In this case, the file ‘m04_s06_e10_angles’ consists of the angular joint measurements for the 10th episode of the 4th movement performed by the 6th subject.

In addition, the data set provides examples of the movements performed in an incorrect manner. The rationale for it is that patients with musculoskeletal injury or constraints are assumed to be unable to, at least initially, perform the prescribed therapy movements in a correct or optimal way during a physical therapy exercise. This portion of the data can serve as a testing set, and it can be used for validation of derived mathematical models for the above described movements. In creating the set of incorrect movements, all of the subjects performed 10 episodes of the 10 movements arbitrarily in a suboptimal manner. The file nomenclature for the incorrect movements follows the same order as for the correct movements, and in addition, all files have an extension _inc, which implies ‘incorrect’. For instance, the file ‘m05_s04_e03_angles_inc’ corresponds to the incorrect 3rd episode performed by the 4th subject for the sit to stand movement.

Description of the joints order for the Vicon and Kinect sensors in the data files, as well as additional description of the position and angular measurements can be found in the paper listed below in the Citation section.

The data is presented in ASCII txt format, with comma delimiter used for separating the data values in the files.

Data Download

Movements (258 MB)

  • Vicon (Positions, Angles) (236 MB)

  • Kinect (Positions, Angles) (32 MB)

Segmented Movements (197 MB)

  • Vicon (Positions, Angles) (173 MB)

  • Kinect (Positions, Angles) (24 MB)

Incorrect Movements (265 MB)

  • Vicon (Positions, Angles) (232 MB)

  • Kinect (Positions, Angles) (33 MB)

Incorrect Segmented Movements (203 MB)

  • Vicon (Positions, Angles) (178 MB)

  • Kinect (Positions, Angles) (25 MB)

Missing Data

The files 'm03_s03_positions/angles' for the Vicon system, related to the inline lunge movement performed by the 3rd subject, are missing. The corresponding data for the same movement recorded with the Kinect sensor are available in the data set.

Reduced Data Set

In our study titled A Deep Learning Framework for Assessing Physical Rehabilitation Exercises we used a reduced version of the original data set, consisting only of segmented angular displacements of the body joints collected with the Vicon motion capture system. In the reduced data set, the episodes of the movements that are not consistent (e.g., the subject used different arm/leg than the other subjects, missing measurement data, movements deviate significantly from the rest of the movements,  etc.) have been removed. In addition, provided also are the movement quality scores which we used for training and evaluting the neural network models in the above paper, as well as the corresponding movement data in csv format. The reduced data set can be downloaded from here (274 MB).

Code for Data Visualization (link)

The code is written in MATLAB and provides visualization of the movement data acquired with the Kinect v2 sensor by using Brekel Pro Body v2 software.


The study for collecting the movements and creating the data set was supported by the Center for Modeling Complex Interactions through NIH Award #P20GM104420 and with additional support from the University of Idaho.

We would like to acknowledge and thank Youngmin Chun for his help with collecting the data, and Christian Williams for his help with organizing the data and segmenting the movements.

We would also like to thank the volunteers who performed the movements for the data set.


If you report results based on the UI-PRMD data set, please cite the following paper:

A. Vakanski, H-p. Jun, D. Paul, and R. Baker, "A data set of human body movements for physical rehabilitation exercises," Data, vol. 3, no. 2, pp. 1–15, 2018. (MDPI Data in Science

Contact or Questions

A. Vakanski at or (208) 757-5422.