Download presentation
Presentation is loading. Please wait.
Published byShannon Kellogg Modified over 9 years ago
1
Copyright Xsens Technologies B.V. 2011; Company confidential Estimating foot parameters Jesper Lansink Rotgerink Supervisors: Dr. ir. Daniel Roetenberg (Xsens) Prof. dr. Stephan A. van Gils (University of Twente)
2
Copyright Xsens Technologies B.V. 2011; Company confidential Xsens Introduce problem Modelling steps Working with real feet Contents
3
Copyright Xsens Technologies B.V. 2011; Company confidential Xsens Spinoff from University of Twente, founded in 2000 Main business: 3D motion tracking One of main products: MVN
4
Copyright Xsens Technologies B.V. 2011; Company confidential MVN: 3D motion tracking Suit with 17 inertial trackers Used in Hollywood, games, rehabilitation Issues with foot Internship assignment: Find out if MVN foot model can be improved using foot sensor data
5
Copyright Xsens Technologies B.V. 2011; Company confidential Placement of foot sensor on foot important for: 1.Estimation of position/velocity 2.Detection of ground contact Different shoes or shoe sizes involve different foot models. What can be ‘seen’ by using only the foot sensor? Motivations
6
Copyright Xsens Technologies B.V. 2011; Company confidential Eventually, try to identify two rotation points: Start: estimate length of a stick Secondly: find radius of a sphere Thirdly: Combine stick and sphere Eventually: Work with real foot Setup
7
Copyright Xsens Technologies B.V. 2011; Company confidential Two reference axes: H is a fixed point. Data from accelerometers and gyroscopes. r is constant in sensor frame. Estimate length of a stick
8
Copyright Xsens Technologies B.V. 2011; Company confidential Leads to linear system from which we can solve r: Least squares fit on the whole data set. Estimate length of a stick
9
Copyright Xsens Technologies B.V. 2011; Company confidential Two formulas for spheres: What is measured by the sensor? We need. Estimate radius of a sphere
10
Copyright Xsens Technologies B.V. 2011; Company confidential For every vector it holds that: Global angular acceleration around x and y axis cause respectively negative y and positive x acceleration: Estimate radius of a sphere
11
Copyright Xsens Technologies B.V. 2011; Company confidential Finally, we come to the linear system: Again solvable by using a least squares fit over the whole dataset. Estimate radius of a sphere
12
Copyright Xsens Technologies B.V. 2011; Company confidential Again two reference axes: We know from previous. Four unknowns. Combination of sphere and stick
13
Copyright Xsens Technologies B.V. 2011; Company confidential Can be written as: Solved by least square fit on whole data set Combination of sphere and stick
14
Copyright Xsens Technologies B.V. 2011; Company confidential Results with test objects
15
Copyright Xsens Technologies B.V. 2011; Company confidential Results with test objects
16
Copyright Xsens Technologies B.V. 2011; Company confidential Results with test objects
17
Copyright Xsens Technologies B.V. 2011; Company confidential Validation to check if our combination method can also ‘detect’ a single sphere or stick: Results with test objects
18
Copyright Xsens Technologies B.V. 2011; Company confidential With respect to time, angular velocity and deflection angle. Sensitivity analysis
19
Copyright Xsens Technologies B.V. 2011; Company confidential Sensitivity analysis With respect to time, angular velocity and deflection angle. Conclusions: Measurements longer then 40 seconds Higher speeds lead to better measurements Deflection angles of 70 degrees and higher are welcome
20
Copyright Xsens Technologies B.V. 2011; Company confidential How does MVN model the foot right now? Input: body height & foot length Average foot Important points Working with real feet
21
Copyright Xsens Technologies B.V. 2011; Company confidential Working with real feet How to deal with different phases of a step?
22
Copyright Xsens Technologies B.V. 2011; Company confidential Tests: Walking Seperate test for ball of the foot Reference data collected with Vicon system Eight infrared cameras that track position of highly reflective markers using triangulation Funny extra: one person wore high heels Working with real feet
23
Copyright Xsens Technologies B.V. 2011; Company confidential Most measurements are okay, up to certain inaccuracy Underestimation Person dependent Results with real feet
24
Copyright Xsens Technologies B.V. 2011; Company confidential Quality of results comparable to previous Advantage of special movement: less standard deviation Results with real feet
25
Copyright Xsens Technologies B.V. 2011; Company confidential Estimates are mostly within a few centimeters from real Script: Can do what we prefer, with little uncertainty With little improvement could be useful to MVN system Especially because height of sensor is determined by shape of shoe and our method can distuinguish. Current MVN cannot. Even high heels worked out fine! Overall conclusion
26
Copyright Xsens Technologies B.V. 2011; Company confidential Underestimations: Deformations of soft tissue Linearization Splitting the data Discussion & further research
27
Copyright Xsens Technologies B.V. 2011; Company confidential Questions? Thanks for your attention
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.