Authors:
Sarvenaz Salehi
1
and
Didier Stricker
2
Affiliations:
1
Daimler Protics, Germany
;
2
German Research Center for Artificial Intelligence (DFKI), Germany
Keyword(s):
Inertial Sensors, Body-IMU Calibration, Body Motion Tracking, Exercise Monitoring.
Abstract:
This work validates the application of a low-cost inertial tracking suit, for strength exercise monitoring. The procedure includes an offline processing for body-IMU calibration, online tracking and identification of lower body motion. We proposed an optimal movement pattern of the body-IMU calibration method from our previous work. Here in order to reproduce real extreme situations, we used data from different types of movements with high acceleration intensity. For such movements, an optimal orientation tracking approach is introduced which requires no accelerometer measurements and it thus minimizes error of existing outliers. The online tracking algorithm is based on an extended Kalman filter(EKF), which estimates the position of upper and lower legs with respect to the pelvis along with hip and knee joint angles. This method benefits from the estimated values in calibration process i.e. joint axes and positions, as well as biomechanical constraints of lower body. Therefore it re
quires no aiding sensors such as magnetometer. The algorithm was evaluated using optical tracker for two types of exercises:squat and abd/adduction which resulted average Root Mean Square Error(RMSE) of 9cm. Additionally, this work presents a personalized exercise identification approach, where an online template matching algorithm is applied and optimised using Zero Velocity Crossing(ZVC) for feature extraction. This results reducing the execution time to 93% and improving the accuracy to 33%.
(More)