Modeling Foot–Floor Interactions during Walking for Normal and Abnormal Gaits

Abnormalities of human walking are critical indicators for the early detection of the risk of trips and falls and of neuromusculoskeletal disorders such as joint impairments, Parkinson’s, and stroke. Understanding the characteristics of dynamic foot–floor interactions during walking (including initial contact, weight translation, and foot clearance) enables the detection of gait abnormalities. Existing studies use various sensing devices to achieve this purpose, including cameras, force or pressure sensors, and wearables. However, they have operational limitations such as minimal visual obstructions (cameras) and small area of coverage (force plates) and require users to carry devices (wearables and pressure insoles), making it challenging for the users to continuously monitor their gait in daily life. In this study, we modeled the dynamic foot–floor interactions during human walking through footstep-induced floor vibrations induced by normal and abnormal gaits. To achieve this, we formulated the problem from the theoretical perspective of structural dynamics and developed experimental analysis by placing vibration sensors on the floor surface to capture footstep-induced structural vibrations. Our approach investigated the mechanism of foot–floor interactions and enabled both force-informed and motion-informed gait analysis using floor vibration data, which has a wide coverage (up to 20 m) and does not require users to carry devices. The main challenge in developing our approach was the complex foot–floor interaction process during walking. Specifically, the interaction area, force magnitude, and force direction vary significantly during walking, leading to difficulty in modeling the entire process explicitly. To overcome this challenge, we first formalized the mechanism of foot–floor interaction into three stages within a gait cycle that has major clinical relevance: initial contact, weight translation during the stance phase, and foot clearance during the swing phase. Then, we formulated the dynamic force characteristics within each stage and derived their impact on floor vibrations. To validate our derivation, we conducted a real-world walking experiment with 10 participants. We collected data from normal and “simulated” abnormal gaits, where participants were instructed to emulate the gait of patients who are commonly observed in clinics under the guidance of medical experts from Stanford Medicine. We observed alignment between theoretical derivation and our experimental data. This established the efficacy of the foot–floor interaction model for gait abnormality detection using footstep-induced floor vibrations. The results also showed significant differences in normal gait among various individuals, posing challenges in detecting abnormal gaits without knowing a person’s normal gait. This motivates future work to robustly detect gait abnormalities across various individuals.