Automatic measurement of physical mobility in Get-Up-and-Go Test using Kinect sensor

Area of Science:

• Gerontology
• Biomedical Engineering
• Computer Science

Background:

• The Timed Up-and-Go (TUG) Test is a standard clinical tool for assessing elderly mobility.
• Objective assessment of gait parameters is crucial for fall risk prediction in older adults.

Purpose of the Study:

• To develop and validate an automated method for analyzing human gait during the Get-Up-and-Go Test using Microsoft Kinect sensor data.
• To classify gait severity and assess fall risk in elderly individuals through machine learning.

Main Methods:

• Utilized Microsoft Kinect sensor to capture human skeleton data during the Get-Up-and-Go Test.
• Extracted gait-related features (steps, duration) and anatomical features (joint angles, limb distances).
• Applied machine learning algorithms, including Bag of Words and Support Vector Machines, for gait classification.

Main Results:

• The automated system successfully extracted relevant gait and anatomical features.
• Machine learning models effectively classified gait severity in a pilot study of 12 elderly subjects (ages 65-90).
• The method demonstrated the ability to discriminate between individuals at high and low risk of falling.

Conclusions:

• Automated analysis of Get-Up-and-Go Test using Kinect sensor data is feasible and effective.
• Extracted gait and anatomical features provide valuable insights into physical mobility and fall risk.
• This technology offers a promising tool for objective assessment and early intervention in geriatric care.