Sampling Rates and Sensor Requirements for Kinematic Assessment During Running Using Foot Mounted IMUs

Abstract

Inertial sensors have the potential to enable in-situ monitoring of athletic performance. They may offer applications in injury prevention and rehabilitation as well as in technique assessment for improved performance. This paper investigates the use of foot worn inertial sensors in order to assess running kinematics. Footwear provides a potential platform for continuous and in-situ monitoring that does not require additional components to be worn by the athlete since inertial sensors are now small enough to be integrated into footwear. These sensors are also inexpensive enough to be accessible to consumers, opening up the possibility of biomechanical assessment not only to elite athletes but also recreational runners. To facilitate widespread adoption by athletes of all types, sensor systems must be as cheap as possible. To achieve this, sensor systems must be engineered with sampling rates that are not unnecessarily high and components that are not over specified. At the same time accuracy requirements must be met. We investigate multiple sensor parameters (sampling rate, acceleration range) and the effects these have on the accuracy of kinematic assessment using foot worn inertial sensors. We find that Extended Kalman Filter based trajectory recovery seems to be little affected by sampling rate until lower than 250 Hz. We investigate impact accelerations using an inertial measurement unit attached to the foot and find that, at 250 Hz, the acceleration signal peaks at up to 70g around heel strike. We also show that the addition of a high range accelerometer improves accuracy of two example metrics that may be useful in gait assessment, maximum foot clearance (FC) and mean step velocity (SV). The 95 % limits of agreement for FC using a ($\pm$ 16g) accelerometer were -4.4 cm to 5.4 cm, this was improved using a high range ($\pm$ 200g) accelerometer. The limits of agreement for FC using the improved system where -2.6 cm to 2.6 cm.