Advancing Ambulatory Gait Stability Assessment with a Small Set of Inertial Sensors
Junhao Zhang is a PhD student in the department Biomedical Signals and Systems. (Co)Promotors are prof.dr.ir. P.H. Veltink from the faculty Electrical Engineering, Mathematics and Computer Science and dr. E.H.F. van Asseldonk from the faculty Engineering Technology
Among older adults and individuals with gait impairments, gait stability problems are often early indicators of an increased risk of falls. Inertial measurement units (IMUs) are characterized by their real-time monitoring capability, convenience, and portability, making them suitable for assessing human gait stability in daily life. Early detection through IMUs allows for the identification of potential stability issues, enabling preventive interventions to reduce fall risk.
In current studies, inverted pendulum-based models are usually used to simplify gait stability assessment. Comparing to the concept of extrapolated center of mass (XCoM) which models the body as a point mass and considers only the linear momentum, the foot placement estimator (FPE) incorporates both linear and rotational momentum. Similar to the XCoM, the FPE gives a point on the ground and from how they are related to the foot placements provides information on whether subjects continue falling, come to stand still or move in the other direction. The inclusion of rotational momentum may provide benefits for assessing gait stability compared to XCoM. Although existing studies have utilized a small set of IMUs to estimate the XCoM, no research has addressed the FPE estimation using a small IMU setup. This task is challenging because it requires estimating both the position and velocity of the CoM and the whole-body angular momentum (WBAM). The goal of this thesis is to propose a solution using a small set of IMUs to estimate the FPE, towards gait stability analysis in daily life. We complete the thesis goal by addressing the following research questions: 1) How can we simplify a full-body model with a reduced set of segments for estimating WBAM? 2) What is the suitable local reference frame for expressing the WBAM and gait stability variables in 3-D cases in daily life conditions? 3) How can we estimate the WBAM and linear momentum, subsequently the FPE with a small set of IMUs? Addressing these three questions leads to the following main findings?
Estimating the sagittal-plane WBAM from a reduced set of segments is possible. Estimating the WBAM based on a small set of IMUs results in an underdetermined system, as not all body segments can be tracked. Therefore, a simplified body model with a reduced number of segments should be used. Our analysis in Chapter 2 indicated that we were able to estimate the sagittal-plane WBAM with respect to the whole-body CoM from a seven-seg model including the head & trunk (HT) and lower limb segments. This finding is expected to have practical implications for the development of the following IMU-based solutions for estimating the WBAM, as they could help reduce the number of required IMUs.
A suitable local reference frame for expressing the WBAM or other biomechanical measures during daily-life tasks has been recommended. For daily activities involving body turns, where anatomical axes rotate with body orientation, using a dynamic local reference frame aligns with the rotating axes, such as one oriented by pelvis heading angle, horizontal CoM velocity, or average angular velocity, can be used for expressing the WBAM. Chapters 3 and 4 showed that reference frame choice significantly affected the WBAM component distribution around different anatomical axes. Applying a low-pass filter with an optimal cut-off frequency reduced their mediolateral oscillations, yielding more stable and anatomically relevant reference frames for precise WBAM estimates. For expressing the WABM or other biomechanical parameters like XCoM and FPE, local reference frames based on the pelvis heading or horizontal CoM velocity are easier to apply and can be derived using a reduced optical marker set or IMUs when whole-body kinematics are unavailable.
A portable solution of using only four IMUs is able to estimate WBAM and linear momentum and subsequently the FPE that considers both the linear and angular momenta of the human body. In Chapter 5, we evaluated whether a single IMU is able to estimate the angular momentum, kinetics, total mechanical energy and its rate of change, of a single rigid body. Inspired by the previous chapters, we proposed a four-IMU-based solution to estimate the WBAM using the simplified model as well as linear momentum, and use these quantities to estimate the FPE. Results from Chapters 6 showed that the estimated whole-body CoM velocity (related to the linear momentum) and WBAM closely matched the reference during straight-line walking and beam walking tasks at both slow and normal speeds. Consequently, the FPE were also estimated with low root mean square errors compared to the reference. The FPE differed from the modified FPE excluding the WBAM only along the anterior-posterior (AP) axis but showed smaller differences along the medial-lateral (ML) axis, indicating a greater influence of WBAM on estimating the FPE along the AP axis than ML axis. Activities that include more intensive ML movement, such as walking tasks with ML perturbations or activities with turns should be included in future study to strengthen the investigation in the added value of incorporating WBAM into FPE.
To conclude, this thesis proposes a portable solution of using only four IMUs to estimate gait stability related variables considering both the linear and angular momentum of the human body. The insights can be considered in the long-term tracking of gait stability in daily life conditions, understanding the contribution of linear and angular momentum to gait stability, and the development of feedback controller for balance assistance.
