Automatic 'Timed-Up and Go' (TUG) test segmentation

Abstract

The Timed-Up and Go test (TUG) is a well-known medical test that is used as an indicator of mental and physical health. I developed the TUG-Segmenter, an automatic segmentation tool that can divide recorded TUG test data into the six main phases of the test: Sitting, Standing-Up, Walking-Forward, Turning, Walking-Back, and Sitting-Down. I created an annotation tool as well that greatly speeds up the creation of ground truth from TUG test data. Using both these tools I was able to evaluate the accuracy of the TUG-Segmenter in terms of the duration of the segmented phases ( 83.4 % accurate ) and the start times of the segmented phases ( 83.6 % accurate). Lastly, I found a 0.3 cm difference for jitteriness and an 8.5 mm/s difference for speed between healthy elderly subjects and healthy young subjects when comparing the features extracted from the individual TUG test phases.