Advancing Inertial Sensors for Performance Quantification: Applications in Balance Rehabilitation and Distance Running

Abstract

Wearable sensors in the form of inertial measurement units (IMUs) enable the unobtrusive quantification of human movement during daily life. As a result, IMUs ubiquitously appear in smart devices such as phones and watches. Substantial research has explored the use of inertial data in performance assessment and feedback systems with applications spanning from biomedical monitoring to athlete training. However, the algorithms required to generate useful insights vary between movement types and user populations. Continued work is required to fully leverage these wearable devices for their potential benefits to human health and performance. This dissertation advances the use of inertial data in two application areas, balance rehabilitation and distance running.

Part I of this dissertation advances IMU-driven feedback systems for use during remote balance rehabilitation programs. Balance rehabilitation is traditionally performed in a clinical setting with the instruction of a physical therapist. However, cost and availability limit access to balance therapy. Technologies improving balance telerehabilitation may therefore improve access to quality care. Inertial sensors can be used to support balance telerehabilitation; training effects can be monitored via automated clinical balance tests, training programs can be automatically or remotely optimized via performance assessments, and the efficacy of training can be improved via feedback.

Two studies herein advance the use of IMUs to support remote balance rehabilitation. In the first study, the effects of IMU-driven terminal (i.e., post-task) visual feedback on sway magnitude and velocity were found to be comparable to those of established concurrent (i.e., real-time) feedback methods during a single session of training. Because IMU-based terminal feedback is possible via a single smartphone, it may further support simple, affordable, and accessible balance training devices. In the second study, self-assessments and IMU-driven kinematic measurements were significantly but imperfectly correlated with expert physical therapist intensity assessments. They may therefore be useful during telerehabilitation when expert visual assessment is difficult. Together, these studies on balance rehabilitation suggest that IMU-based systems may support high-quality home-based care.

Part II of this dissertation advances the use of IMUs for “real-world” distance running assessment. While analyses of running gait have traditionally been performed in a laboratory, running biomechanics, physiology, and psychology all differ between laboratory and “real-world” settings. IMU-based assessments of “real-world” distance running may therefore capture “real-world” performance determinants, injury etiology, and training or intervention effects. Additionally, IMU-based systems may be less expensive and more accessible than traditional laboratory-based systems. However, additional research is required to expand the set of metrics available via IMUs and to establish relationships between inertial data and important aspects of running performance.

Two studies in the dissertation advance the use of IMUs to assess “real-world” distance running. In the first study, an error-state Kalman filter (ErKF) algorithm for the estimation of three-dimensional lower-body running kinematics resulted in joint angle estimates of comparable accuracies to existing methods. ErKF algorithms may also support adaptation to experimental design (e.g., speed, running surface) and should therefore be further advanced for kinematic estimation during “real-world” running. In the second study, numerous biomechanical measures including peak acceleration attenuation and stability were associated with positive running experiences during long, outdoor training runs. These relationships should be additionally explored as means of supporting positive running experiences. Together, these studies on distance running demonstrate that wearable sensors can provide meaningful measures of performance during “real-world” running.