Automated Segmentation of the Timed-Up-And-Go Test Using Millimetre-Wave Radar Sensors for Mobility Assessment

The Timed Up and Go (TUG) test is a widely adopted clinical assessment for evaluating functional mobility and identifying falls risk. This study presents a method for automating the segmentation of the TUG test using millimetre-wave radar sensors, with the aim of enabling real-time, non-contact mobility assessment. By leveraging temporal variations in radar signals, a segmentation algorithm was developed to delineate key movement phases, including sit-to-stand transitions, walking forwards, turning, walking back, and sitting down. The proposed approach facilitates the extraction of digital biomarkers that may provide clinically relevant insights into mobility patterns and health status. Proof-Of-concept experiments were conducted in controlled environments to evaluate segmentation accuracy and action duration prediction. The model segmented the sub-tasks with mean absolute percentage error values below 5% and correlation coefficients above 0.96 for most actions. Sub-Task durations were estimated with mean absolute errors ranging from 0.16 to 0.23 seconds, and correlation of 0.9978 for the total TUG test duration. These initial findings support the feasibility of using radar-based systems for unobtrusive and scalable mobility monitoring in both clinical and residential settings.