Use of Nonlinear Analysis Methods for Evaluating IMU Data of Bilateral Jump Landing Tasks

The use of nonlinear analysis methods provides new information when evaluating linear acceleration and angular velocity from a system with Inertial Measurement Unit (IMU) recording. This information is used as additional input to improve the estimation of the angular displacements in a neural network model. The measurements were performed on 24 participants (18 males and 6 females of an average age of $22.6\pm \ 2.6$ years old, average height of $172.6\pm 10.3$ cm, and an average weight of $72.2\pm 16.02\ \text{kg})$ during bilateral jump landing tasks. In order to assess the differences between IMU estimated angle and the gold standard, data obtained from Qualysis optical Mocap (Qualisys AB, Göteborg, Sweden) and Delsys inertial measurement systems (Delsys Inc., Boston, MA, USA) were used for measurements during bilateral jump landing tasks. A total of 8 IMU sensors were placed on the sternum, L5, bilateral thighs, shanks, and foot. The thigh and shank sensors were placed on the middle of each thigh and shank along the anterior-posterior axis (middle thigh and middle shank) while the foot sensors were placed on the dorsal surface of the foot. Thirty retroreflective markers were placed on the pelvis and bilateral thigh, shanks, and foot to form a 7-linkage lower extremity model. Static calibration on each of the participants was performed during standing with anatomical position to define the neutral joint angle at bilateral hip, knee, and ankle. For quantification purposes, the Hurst exponent, Lyapunov exponent, approximate entropy, and multiscale sample entropy were used. The results suggest that when evaluating the placement of IMU on the shank and thigh to determine the knee angle, the Hurst exponent is capable of best distinguishing individual axes based on linear acceleration and angular velocity.