Date of Award

2019

Degree Type

Thesis

Degree Name

Master of Science in Mechanical Engineering

Department

Washkewicz College of Engineering

First Advisor

Bogert, Antonie van den

Subject Headings

Biomechanics, Design, Kinesiology, Mechanical Engineering, Sports Medicine, Technology

Abstract

The purpose of this research is to develop a system of motion capture based on skin-applied strain sensors. These elastic sensors are of interest because they can be applied to the body without restricting motion and are well suited to operate in more practical environments, such as sports fields, gymnasiums, and outdoor areas. This combination is currently not available in the field of motion capture. The current issues with strain sensor motion capture technology is the accurate is not sufficient for motion analysis and axial rotation monitoring of joints is not available. This project will build and test a sensor arrangement designed to measure axial joint rotation and a calibration that compensates for crosstalk from other joint motions.

An arrangement of four strain sensors was created to capture hip and knee motion indirectly through a geometric relationship. Sensors were arranged around the hip and knee with compression pants that emulate the pressure sensation of kinesiology tape. This pressure is desirable for high level athletes are comfortable with this feeling, meaning most wearers would likely agree. This prototype was tested on six participants of varying height with Institutional Review Board approval and was referenced against a passive marker, visual motion capture system with ten cameras.

The test results show the geometric calibration with crosstalk compensation is the most successful general calibration. The overall root-mean-square error of the hip flexion, hip abduction, hip rotation, and knee flexion measurements were 4.6±1.2° (ρ = 0.95), 4.7±1.5° (ρ = 0.82), 6.7±2.0° (ρ = 0.89), and 6.2±1.3° (ρ = 0.96) respectively, compared to a commercial xSens system with 5.7±2.1° (ρ = 0.99), 4.1±2.0° (ρ = 0.91), 6.5±2.8° (ρ = 0.68), and 4.4±2.0° (ρ = 0.99). The geometric calibration without crosstalk compensation tends to miss the relation of the data but may be sufficient for small ranges of motion. Specifically, axial rotation sensing capacity was shown to be important for the accuracy of other sensor’s angle readings. The gaussian processes regression (GPR) tended to overfit the calibration data. In conclusion, the geometric calibration with crosstalk compensation created the most successful, stable, and general calibration. This testing was performed with a $200 prototype and produced results comparable to a $20,000 commercial system.

COinS