Detection of gait cycles in treadmill walking using a Kinect

Area of Science:

• Biomechanics
• Human Motion Analysis
• Medical Technology

Background:

• Treadmill walking is essential for gait analysis in clinical settings.
• Depth cameras like Kinect offer low-cost gait analysis potential.
• Kinect's accuracy in detecting foot-ground contact (heel-strike) is limited due to depth image interpretation challenges.

Purpose of the Study:

• To develop and validate an indirect method for estimating heel-strike events during treadmill walking using Kinect.
• To assess the accuracy of this new method compared to traditional optoelectronic systems.
• To improve the reliability of gait analysis in routine outpatient clinics.

Main Methods:

• Recorded motion of 11 healthy subjects walking on a treadmill using an optoelectronic system and Kinect.
• Assumed heel-strike events could be indirectly estimated by analyzing extreme values in the longitudinal distance between knee joints.
• Used anthropometric data to locate knee positions in Kinect's depth maps.

Main Results:

• The Kinect-based knee-tracking method yielded heel-strike estimation errors of 17±24ms.
• Mean gait cycle duration errors were minimal (0±12ms) with the Kinect approach.
• The proposed method demonstrated improved accuracy compared to previous Kinect-based gait studies.

Conclusions:

• Indirect estimation of heel-strike events via knee joint tracking is a viable alternative for Kinect-based gait analysis.
• This method shows potential for enhancing the accuracy and clinical utility of low-cost gait analysis systems.
• Further validation could lead to more widespread adoption in routine outpatient gait assessments.