Automated Timed Up and Go test segmentation via pose detection

The Timed Up and Go test is commonly used for assessing gait, balance and function in people with limited mobility or low functional mobility. The test consists in measuring the time it takes for a subject to stand up, walk 3 meters, come back and sit down. Computer vision and pose estimation allows measuring the subject's performance more reliably than a human observer and to extract additional parameters that are potentially more sensitive for the assessment of symptoms than the total duration. We show an algorithmic approach for multi-camera settings and present accuracy results obtained in a pilot study with 13 older adults, establishing the methodology for larger studies.