Inertial Sensor-based Knee Angle Estimation for Gait Analysis Using the Ant Colony Algorithm to Find the Optimal Parameters for Kalman Filter

Abstract

This paper describes a new method for human joint angle calculation based on inertial measurement data. By exploiting kinematic constraints based on anatomical knowledge, a method is developed that avoids the assumption that the sensors must be mounted in certain orientations towards the body segments. The joint axis is identified by calculating the inertial measurement units' (IMUs') quaternion, and then the joint angle is calculated from segment acceleration and angular velocity data according to the joint axis, respectively. As a tool for sensor fusion, a Kalman filter is used to combine both angles. There are two parameters in the Kalman filter that are always assigned to experiential values. In this paper, a standard subject experiment is designed to optimize the parameter combination in two dimensions by using the ant colony algorithm. The results showed that the root-mean-square error (RMSE) between the estimated rotation angles and the programmed angles was less than 1 degree after the optimal parameters' values were found by the ant colony algorithm.