Recognizing and categorizing qualities of human movement using sensors

Replies to This Discussion

Permalink Reply by Marco Fusion on March 8, 2010 at 7:31pm

You probably want to look at flex sensors. As it turns out, last week I posted a tutorial on setting up a wireless flex sensor, for input into Isadora or Max/MSP/Jitter, etc.:
http://fusion-artist.com/XBee/XBeeTutorial.htm
The flex sensor can measure the angle of a joint, e.g. the elbow. So by tracking how fast that angle changes (in your Isadora or Max patch), you could guess at whether it's a punch (fast), poke (medium) or slow (a pat.) For more accuracy you could use multiple flex sensors, or integrate camera input tracking through something like EyesWeb, though that adds complexity.
Of course there's no perfect way to ascertain variables like "intention" with sensors. There are qualities that are far more subtle than what can be picked up by sensors. You might be interested in OSU's "Synchronous Objects" project, where they spent years tracking dozens of simultaneous Laban variables for a William Forsythe piece:
http://synchronousobjects.osu.edu/
To do that, they needed to do manual study of each dancer's movement, and even integrate interviews from the dancers themselves to determine cues and intentions.

Permalink Reply by Diego Maranan on March 8, 2010 at 10:33pm

Thanks, Marco. I'm using XBee accelerometers for this project and probably using a neural network to train the system, although your post raises the intriguing possibility of validating our system (which uses accelerometer data) using another kind of data (angle changes). The system I'm working on was used to an extent for a remount of Trisha Brown's Astral Convertible at the University of Illinois at Urbana-Champagne.

I saw Synchronous Objects before, but I didn't realize that they were tracking Laban-related things. I will go back and look at it; thanks for the tip!

Have you used your system for a performance?

Incidentally, here are some recent work down on Laban-based recognition.
https://www.zotero.org/groups/laban_effort_researchmeaning_from_mot...

Permalink Reply by Diego Maranan on March 9, 2010 at 2:48am

I took a look again at the Synchronous Objects website, and the only data I saw them track and use was choreographic data such as cues, positions of where the bodies were in space and what they were doing, but I suspect that one needn't use LMA to do any of that. And they certainly didn't seem to record and keep track of movement qualities in a "semantic" or even systematic way. Am I missing something here? I will go check publications to see if any of the people working on SO mention something.

Permalink Reply by Lali Krotoszynski on April 20, 2010 at 8:05pm

Hello Diego,
I am a Brazilian choreographer and I am presently working on a system for an installation which will translate retrieved data from visitors into projected creatures moving accordingly to Laban's effort actions.
I would like to know more about your project. Maybe we can exchange info...
Cheers,
Lali

Permalink Reply by Diego Maranan on April 20, 2010 at 10:05pm

Hi Lali,

At present I am working on building a system that was implemented in the past. I am working with some researchers from my uni, Simon Fraser University, and from the University of Illinois at Urbana Champaign. It's still very incomplete, so I have nothing yet to show you. However, please drop me a line at dmaranan at upou dot edu dot ph, and we can keep in touch.

Also, LMA-based recognition systems have been implemented by other researchers, most notably by Norman Badler's group at the University of Pennsylvania. A few researchers in Japan have also worked on this. If you want more information, I have a list of published papers on http://www.zotero.org/groups/motion_frameworks_for_physical_express.... Maybe you can get in touch with the authors of some of the publications.

Cheers,
Diego

Permalink Reply by Vangelis Lympouridis on October 11, 2010 at 4:42pm

Hello,
that is an interesting approach. I believe that the efficiency of the system you aim to create and propose as a tool will affect the translation of creative intentions to artistic forms. Accelerometers are nice but I think very limited representing the qualities you are after. I have used a whole body sensor based mocap system to do my research. Although very promising these systems are still under development. The best sensor based mocap of the self system is Xsens MVN from Moven. Many others are now developed at research labs. Flex sensors are very inefficient (you will find out directly if you try to use them). It is hard to carry on these projects as there are many technical issues that engineers have to first resolve. Good luck. Let me know if you would like more info on my work. v.