Hi Jack
We are currently working on implementing a general method for handling of
kinematically overdetermined systems, like the marker driven models.
This work is based on the work by Michael Skipper Andersen, please see this
webcast for details (you will need to log into the user area to see this)
<http://www.anybodytech.com/199.0.html>
http://www.anybodytech.com/199.0.html Kinematic analysis of
overdetermined-systems by Michael Skipper Andersen 22.02.2007
This method lets you optimize the joint angles, local marker positions, and
segment lengths to give the best possible match between measures markers and
model markers attached to the bones, throughout the trial. So in other words
it automatically scale the segments length and let modify the initially
given marker locations to match the experiment.
This method has been implemented for the lower extremity in a hard coded
external application “gaitapplication2”. This application contains a
hardcoded kinematic leg model that has the exact same topology as the
current leg model in the repository. The GaitUniMiami model is an example on
how the external gaitapplication2 can be used together with the leg model in
AnyBody for optimizing, motion, marker location and segment lengths.
As written above, this technique is currently being implemented as a general
feature in the system; this will in time allow the user to define these
kinds of optimization problems in a general manner. So it can be applied to
for all kinds of body models.
The implementation is ongoing and on track, but there are no release date or
version number settled for this yet, it is a large task, since it involves
changes in the kinematic engine.
If you need a solution before this becomes available and can do the
optimization outside the AnyBody, there are ways to output and input data.
We will of course try to help you in this process.
We will be present at the NACOB conference in August, so if you attend this
conference there will be a chance to discuss these issues with us.
Please ask again if you have further questions.
Best regards
Søren, AnyBody Support
From: anyscript@yahoogroups.com [mailto:anyscript@yahoogroups.com] On Behalf
Of Jack Winters
Sent: 08 July 2008 16:00
To: anyscript@yahoogroups.com
Subject: [AnyScript] Automatic MoCap Initialization for Torso-Arm?
Hi Sylvain (or ?),
There has been a past discussion on using Mocap data as related to
the challenge initialization (trying to match the body and collected
markers). We have been trying to use the “manual” method for the
torso-arm model (our movement task is reaching to various regions of
the workspace and applying forces in 3D directions; participants are
persons with disabilities) with limited success. It takes a very
long time to iterate to a working set of initial position (assuming
scaling is OK), prohibitively so for a group of subjects going to 27
locations within their workspace. We have been thinking about using
optimization methods to do this, as obviously this is a process best
done based on all of the marker data, and the “automatic” way that
exists for the lower extremity is the way to go.
Two questions:
-
Are there plans to implement the automatic method (which is
really the only way that makes sense) for the rest of the body?
-
If not, would you be willing to work with us to make this happen,
or give some advise on the best way without having to do it with our
own optimization approach that passes data to/from Anybody?
Also, in a separate note I’ll mention a few questions related to
scaling.
Jack
[Non-text portions of this message have been removed]