Accuracy Improvement on the Measurement of Human-Joint Angles.

A measurement technique that decreases the root mean square error (RMSE) of measurements of human-joint angles using a personal wireless sensor network is reported. Its operation is based on virtual rotations of wireless sensors worn by the user, and it focuses on the arm, whose position is measured on 5 degree of freedom (DOF). The wireless sensors use inertial magnetic units that measure the alignment of the arm with the earth's gravity and magnetic fields. Due to the biomechanical properties of human tissue (e.g., skin's elasticity), the sensors' orientation is shifted, and this shift affects the accuracy of measurements. In the proposed technique, the change of orientation is first modeled from linear regressions of data collected from 15 participants at different arm positions. Then, out of eight body indices measured with dual-energy X-ray absorptiometry, the percentage of body fat is found to have the greatest correlation with the rate of change in sensors' orientation. This finding enables us to estimate the change in sensors' orientation from the user's body fat percentage. Finally, an algorithm virtually rotates the sensors using quaternion theory with the objective of reducing the error. The proposed technique is validated with experiments on five different participants. In the DOF, whose error decreased the most, the RMSE decreased from 2.20(°) to 0.87(°). This is an improvement of 60%, and in the DOF whose error decreased the least, the RMSE decreased from 1.64(°) to 1.37(°). This is an improvement of 16%. On an average, the RMSE improved by 44%.