Forums > MaxMSP

Wii Gestural Recognition

February 26, 2007 | 7:56 pm

I’ve been trying to figure this out for hours but have come to a roadblock…perhaps there is something simple I am just overlooking?

I am trying to set up gesture recognition so that I can record movement with my Wiimote, and then use those movements to trigger sounds.

What’s the best way to accomplish this? For now, i’ll get back to fiddling and trying to figure it out…


February 28, 2007 | 11:53 am

hi,
i’ve tried to reply by sending a mail to the list but it seems it didn’t work.
anyway…
where’s your real problem? i mean, is it :

-how do i record movements?
-how do i recognize movements from the recorded ones?

i think the first is easier than the second.
the first is ‘just’ a computer problem, i mean, all you need is dev skill.
the second is a bit harder, since you’ve to use data mining methods, you will have to choose the right data representation, the right distance computation, etc…

i’ll receive a wiimote soon, and i want to do exactly the same thing, i’ll keep you informed…

bye

GE


March 1, 2007 | 4:51 am

I’m actually having problems even from the beginning– recording movements… I think I have managed to figure a little out and get some recorded movements working using the MNM externals (part of FTM)…but I haven’t yet figured out how to use it like I need to…

If I am conceptualizing this right, in order do specific movement with the wiimote, (because depending on how fast you do the movement the variables are different), it would probably be best to record the movement to generate some kind of visual line/curve? If I use the raw data, speed of the movement will change the results right? If the data (numbers) from the movement draw some kind of shape, maybe I could use that shape recorded to trigger the sounds…

Maybe I am just thinking way too much and there is an easier and better way? Any insight into a more efficient route and/or how to get there is appreciated. Thanks a lot.

Jordan


March 1, 2007 | 4:52 am

I’m actually having problems even from the beginning– recording movements… I think I have managed to figure a little out and get some recorded movements working using the MNM externals (part of FTM)…but I haven’t yet figured out how to use it like I need to…

If I am conceptualizing this right, in order do specific movement with the wiimote, (because depending on how fast you do the movement the variables are different), it would probably be best to record the movement to generate some kind of visual line/curve? If I use the raw data, speed of the movement will change the results right? If the data (numbers) from the movement draw some kind of shape, maybe I could use that shape recorded to trigger the sounds…

Maybe I am just thinking way too much and there is an easier and better way? Any insight into a more efficient route and/or how to get there is appreciated. Thanks a lot.

Jordan


March 1, 2007 | 5:00 am

have you looked at techniques like autocorrelation? I’m sure a good
number of the audio domain analysis algorithms that try to understand
if an incoming signal matches a database of stored samples will be
relevant to what you’re trying to do with the wii.

wes


March 1, 2007 | 5:07 am

hey wes,

i had never heard of autocorrelation but after briefly looking it up it seems interesting…i need to do some more research on it, but any information you have or anyone else is appreciated… hmm, at least i have another place to look


March 1, 2007 | 5:25 am

basically you’ll store a sequence of values representing a gesture and
compare that sequence to an incoming stream of numbers. It’s very
similar to convolution. Here’s a nice little applet that shows what
autocorrelation does:
http://cnyack.homestead.com/files/aconv/convau1.htm . Notice how you
get peaks in the signal when the input stream and stored gesture
align.

wes

On 2/28/07, Jordan Hochenbaum wrote:
>
> hey wes,
>
> i had never heard of autocorrelation but after briefly looking it up it seems interesting…i need to do some more research on it, but any information you have or anyone else is appreciated… hmm, at least i have another place to look
>


March 2, 2007 | 10:25 pm

that is very interesting…im assuming that would directly apply to taking the input stream of the wii data and basically matching it up with stored information– which would do exactly what i need. Do you know of any externals that do something like this or would this be my first java-scripting project? hehe


July 4, 2008 | 1:24 pm

Hi there,

For my graduation project I created a gesture recognition patch in Max MSP with the help of Akamatsu’s wiiremote object and my own externals.

It kinda works the same as liveMove from AiLive.

http://www.youtube.com/watch?v=5ffvSMWD2Mk

I think this can be very interesting for creating your own live performance/composing environments. Anyone interested in using this just let me know! I hope I can make a stable, easy to use version and make it available around August.

cheers,

Tim



kjg
July 6, 2008 | 3:16 pm

Hi Tim,

I would be very interested to see your patch, and I am pretty sure I am not the only one :)

Please post when you are ready.

regards,
kjg


July 6, 2008 | 6:36 pm

did you try the gesture follower that comes with mnm?

http://ftm.ircam.fr/index.php/Gesture_Follower

oli


July 6, 2008 | 7:52 pm

Quote: oli larkin wrote on Sun, 06 July 2008 12:36
—————————————————-
> did you try the gesture follower that comes with mnm?
>
> http://ftm.ircam.fr/index.php/Gesture_Follower
>
> oli
—————————————————-

No I haven’t till just now but it also works pretty good!


September 19, 2009 | 12:49 pm

Hi All!

I’m trying to develop a patch in max to recognize wii gestures,
but it has been very difficult since I’m not getting very accuracy in the results.
I think I’m not treating right the raw data from wii (osculator-osc).
So I was trying to translate this java code (www.wiigee.org) to mxj
but it has been hard since I’m a novice in java.
Can anyone help? I would be very much grateful, and I think the community also.

Best Regards,
DC


Viewing 13 posts - 1 through 13 (of 13 total)