Elsevier

Applied Ergonomics

Volume 43, Issue 4, July 2012, Pages 645-649
Applied Ergonomics

Evaluation of the Kinect™ sensor for 3-D kinematic measurement in the workplace

https://doi.org/10.1016/j.apergo.2011.09.011Get rights and content

Abstract

Recording posture and movement is important for determining risk of musculoskeletal injury in the workplace, but existing motion capture systems are not suited for field work. Estimates of the 3-D relative positions of four 0.10 m cubes from the Kinect were compared to estimates from a Vicon motion capture system to determine whether the hardware sensing components were sensitive enough to be used as a portable 3-D motion capture system for workplace ergonomic assessments. The root-mean-squared errors (SD) were 0.0065 m (0.0048 m), 0.0109 m (0.0059 m), 0.0057 m (0.0042 m) in the x, y and z directions (with x axis to the right, y axis away from the sensor and z axis upwards). These data were collected over a range of 1.0–3.0 m from the device covering a field of view of 54.0 degrees horizontally and 39.1 degrees vertically. Requirements for software, hardware and subject preparation were also considered to determine the usability of the Kinect in the field.

Highlights

► The suitability of the Kinect system to be used for 3D motion capture is examined. ► Accuracy of the Kinect is compared to a Vicon system. ► Required software, hardware and subject preparation are also considered.

Section snippets

Background

Recording posture and movement is important for determining risk of musculoskeletal injury in the workplace (Vieira and Kumar, 2004). Existing lab-based three-dimensional (3-D) motion capture systems are of limited use for performing ergonomic assessments in the field. Both active (e.g. NDI; Waterloo, Ontario) and passive (e.g. Vicon Motion Systems; Los Angeles, California) video based systems are difficult to use in real-world applications due to complexity, bulk and space requirements (Best

Materials and methods

A Kinect sensor was connected to the USB port of a Dell XPS600 desktop computer with 2GB ram, 2.8 GHz dual core Intel Pentium processor running the Ubuntu 10.10 operating system. Data from the Kinect was collected using Kinect RGB Demo v0.3.0 software (Burrus, 2011). The software included drivers for communicating with the sensor via USB, a utility for calibrating the sensor as well as the ability to save the 3-D image data and later export the data from each depth map to a text file. The

Results and discussion

Stem plots of the RMS errors of the 104 data points collected by the Vicon and Kinect systems are shown in Fig. 5. RMS errors in the x and y directions are shown with respect to the x and y position in the capture volume. RMS errors in the z direction are shown with respect to their y and z positions in the capture volume.

Conclusion

The Kinect was able to capture the relative 3-D coordinates of markers with RMS errors (SD) of 0.0065 m (0.0048 m), 0.0109 m (0.0059 m), 0.0057 m (0.0042 m) in the x, y and z directions, respectively, using the Vicon system as a gold standard. The system provided this accuracy over the range of 1.0 m to 3.0 m from the camera with an effective field of view of 54.0° horizontal and 39.1° vertical. Therefore, with a small amount of further development, the Kinect may provide investigators with a portable

Acknowledgements

The author would like to thank Ms. Tonya Martin for her help with data collection and editing, Dr. Karl Zabjek for the use of his biomechanics lab and Vicon motion capture system as well as Mr. Kaveh Momen for his expertise with the Ubuntu operating system. The author also acknowledges the support of Toronto Rehabilitation Institute who receives funding under the Provincial Rehabilitation Research Program from the Ministry of Health and Long-Term Care in Ontario (MOHLTC Grant # 06036).

References (16)

There are more references available in the full text version of this article.

Cited by (358)

View all citing articles on Scopus
View full text