Kinect as a tool for gait analysis: validation of a real-time joint extraction algorithm working in side view

Area of Science:

• Biomechanics
• Rehabilitation Engineering
• Computer Vision

Background:

• The Microsoft Kinect sensor is a valuable tool for gait analysis.
• A limitation of Kinect is its inability to extract joint data from side-view body images.
• This restricts its application in various relevant fields.

Purpose of the Study:

• To present an algorithm for locating and estimating joint trajectories from Kinect side-view depth data.
• To apply this algorithm for objective scoring of the "Get Up and Go Test" in gait analysis.
• To offer a computationally efficient alternative to machine learning methods.

Main Methods:

• The algorithm uses anthropometric models to identify joint positions from Kinect depth data.
• It processes depth-data streams to estimate trajectories of up to six joints.
• The method avoids complex computations, requiring fewer resources than machine learning.

Main Results:

• The algorithm successfully extracts joint position data from side-view depth images.
• The extracted joint trajectories demonstrate high adherence to reference curves.
• Performance surpasses that of native Microsoft Kinect and OpenNI SDK skeleton tracking.

Conclusions:

• The proposed algorithm enhances Kinect's utility for gait analysis, particularly in rehabilitation.
• It provides a reliable and computationally efficient method for joint trajectory extraction.
• This advancement facilitates more objective assessments like the "Get Up and Go Test".