Age-Dependent Mobility Decline Analysis Through Sequential Foot Motion Reproduction

Abstract

Most elderly people are experiencing mobility declines due to aging, which often causes falls and injuries. The timed-up and go is a widely used clinical test for mobility evaluation and analysis that consists of sequential motions of gait and U-turn, in addition to standup and sit from a chair. Compared to such a clinical evaluation tool, smart insoles, when integrating with sensors and data processing techniques, as an inexpensive daily assessment alternative, can help the elderly monitor their mobility at home. In this letter, we use accelerometer-based monitoring insoles (MONI) developed in our lab to collect and process sequential foot motion, i.e., gait and U-turn, which is a simple setup without the need for a chair. By integrating with the label model framework (LMF), MONI can monitor and detect age-dependent mobility changes of elderlies with high accuracy (96.13% for 22 subjects). The results are generated by collecting sequential motion data from six elderly people and 16 young adults using MONI during a round trip 10-ft walking with one U-turn. The LMF with weak supervision can robustly reproduce the sequential motions and then extract 21 temporal motion features from the reproduction results. By ranking these features, we can identify their relevance in representing age-dependent mobility decline and the importance of sequential effects for mobility analysis. Errors associated with sequential motion reproduction are analyzed, and the applicability of our proposed method is proven.