Land and Underwater Gait Analysis Using Wearable IMU

Abstract

Walking underwater reduces joint impacts, enhances stability and lowers the net body weight of the patient during rehabilitation. It is a recent rehabilitation method and few suitable methods exist to study underwater gait kinematics. We propose an underwater inertial measurement (IMU) system analogous to those used in land-based rehabilitation to investigate gait kinematics. The objective of this study was to test and validate the proposed system in two human trials by evaluating the knee angle during the gait. In the first trial, a three-way performance analysis was carried out between the IMU, optoelectronic and motion-capture systems in a traditional rehabilitation setting on land. In the second trial, the proposed underwater IMU is compared with camera-based motion-capture both inside and outside the water environment, using the same subjects in both phases of the trial. This allows for an evaluation of the walking gait in air and underwater as well as a cross-comparison of IMU-based knee angle estimates before and after Gaussian Process Regression. The major finding of this work is that the proposed underwater wearable IMU system provides reliable and repeatable measurements of the knee angle during the gait, both in air and underwater. Index Terms-gait analysis, inertial measurement unit (IMU), kinematics, rehabilitation, underwater, optoelectronic tracking, motion-capture I. INTRODUCTIONW ATER provides a nearly ideal environment for physical rehabilitation. This is due to the additional forces acting on the submerged body, primarily caused by the dynamic pressure and drag. These hydrodynamic forces reduce the net body weight, lower joint loading and provide enhanced physical support and posture stabilization [1], [2]. Walking rehabilitation in water, shown in Figure 1, affects the muscular skeletal system [2], by reducing fatigue and pain, improving the physical recovery rate as well as joint

Wearable technologies for monitoring aquatic exercises: A systematic review

“…Table 1 summarizes the characteristics and demographics of the included articles, showing that the majority of studies were published between 2017, 31 – 38 2019 39 – 42 and 2020 43 – 46 and no articles were found before 2014. Most of the articles 31 – 33 , 35 – 37 , 42 , 44 – 51 involved healthy adults or elderly subjects, 41 , 43 while the remaining investigated anterior cruciate ligament injury, 34 , 52 incomplete spinal cord injury 39 , 40 , 53 and chronic anterior knee pain. 38 Three studies involved a healthy control group, 34 , 38 , 39 and one 52 exploited previously published data of a healthy reference group.…”

“…This variety is also reflected in the adopted protocols, exercises and evaluation metrics. Gait analysis is performed on dry land and underwater in nine studies, 34 , 42 , 43 , 45 – 49 , 51 , 52 following in four 34 , 43 , 48 , 52 cases the Outwalk protocol. 54 In all of these works, as well as in studies evaluating running on a treadmill, 32 , 33 the measured outcomes are focused on temporo-spatial parameters, joint kinematics and range of motion.…”

“… 34 , 37 , 38 , 41 – 43 , 48 , 52 , 53 Five articles included motion capture and optoelectronic system as support, 45 , 46 , 51 reference 32 or as gold standard to validate the inertial sensors. 44 , 49 Force plates were used by Marinho-Buzelli et al 35 , 36 , 39 , 40 to estimate the centre of pressure parameters and sway, and by Pacini Panebianco et al 46 and Chien et al 50 to determine ground reaction forces. Lastly, Chien et al, 31 accompanied the accelerometer data with electromyography for the investigation of muscular contraction and near-infrared spectroscopy to estimate the tissue saturation index.…”

Foot Position Measurement during Assistive Motion for Sit-to-Stand Using a Single Inertial Sensor and Shoe-Type Force Sensors

“…Thus, this study compared the outcomes of artificial neural network (ANN), Gaussian process, k-nearest neighbor (kNN), linear regression, and support vector regression (SVR) for the proposed method. These five machine learning algorithms were selected given their frequent use for wearable systems in previous studies [ 32 , 33 , 34 , 35 , 36 ]. In addition, ANN, Gaussian process, kNN, and SVR were expected to build regression models for nonlinear and complex data distribution [ 37 , 38 , 39 , 40 ].…”

Estimation of Lower Extremity Muscle Activity in Gait Using the Wearable Inertial Measurement Units and Neural Network

“…The sensors are small and light and can capture human motion outside a laboratory environment. Several studies have demonstrated the accuracy and reliability of the IMU for gait analysis [ 1 , 2 ].…”