Kalman filtering with accelerometers

[Eirik Arthur Blekesaune]

Hi,
I'm currently working on an interactive danceperformance, using  
accelerometers. I'm now trying to implement some kind of position  
measurements with the signals from the accelerometers. During my  
googling I've found this article: http://www.gamasutra.com/view/feature/1494/wheres_the_wiimote_using_kalman_.php?page=1

I wondered if someone has done anthing with processing accelerometer  
data using SuperCollider. Trying here to find differrent ways of  
extracting data from the sensor signals. Maybe even using Kalman  
filtering. I've been using  Nescivi's SenseWorld, which is really nice.

Has anyone else done some implementations using sensordata processing,  
and maybe with accelerometers too?

-Eirik Arthur Blekesaune
_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[nescivi]

On Wednesday 07 May 2008 07:28:50 Eirik Blekesaune wrote:

> Hi,
> I'm currently working on an interactive danceperformance, using
> accelerometers. I'm now trying to implement some kind of position
> measurements with the signals from the accelerometers. During my
> googling I've found this article:
> http://www.gamasutra.com/view/feature/1494/wheres_the_wiimote_using_kalman_
>.php?page=1
>
> I wondered if someone has done anthing with processing accelerometer
> data using SuperCollider. Trying here to find differrent ways of
> extracting data from the sensor signals. Maybe even using Kalman
> filtering. I've been using  Nescivi's SenseWorld, which is really nice.

Thanks. Glad to hear that they are being used :)
I am interested in any suggestions, since I am still developing them further.

the SenseWorld classes come from working on an interactive dance piece called
Schwelle, which we have been performing a couple of times now.
We have written papers about it for last year's NIME and ICMC, so you may want
to check these out for some background information, and references to other
projects.

Getting actual position measurements with accelero's is tricky, as you lack
two integration constants, when you integrate from acceleration to speed to
position.
I'll take a closer look at the article you posted, it shouldn't be possible to
implement that in SC. Did you have a go at that already?


sincerely,
Marije
_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[Alessandro Fogar]

Hi,

2008/5/8 nescivi <nescivi@...>:
> We have written papers about it for last year's NIME and ICMC, so you may want
> to check these out for some background information, and references to other
> projects.

are these paper downloadable?

All the best

--
Alessandro Fogar

http://www.fogar.it
_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[nescivi]

On Thursday 08 May 2008 03:56:35 Alessandro Fogar wrote:
> Hi,
>
> 2008/5/8 nescivi <nescivi@...>:
> > We have written papers about it for last year's NIME and ICMC, so you may
> > want to check these out for some background information, and references
> > to other projects.
>
> are these paper downloadable?

The NIME one is for sure: http://www.nime.org/2007/proc/nime2007_178.pdf
this one has the most technical descriptions.

The ICMC one is not, but I can send it to anyone interested upon request.

sincerely,
Marije
_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[Scott Wilson-8]

Marije,

are the SenseWorld classes available?

S.

On 8 May 2008, at 08:49, nescivi wrote:

> On Wednesday 07 May 2008 07:28:50 Eirik Blekesaune wrote:
>> Hi,
>> I'm currently working on an interactive danceperformance, using
>> accelerometers. I'm now trying to implement some kind of position
>> measurements with the signals from the accelerometers. During my
>> googling I've found this article:
>> http://www.gamasutra.com/view/feature/1494/wheres_the_wiimote_using_kalman_
>> .php?page=1
>>
>> I wondered if someone has done anthing with processing accelerometer
>> data using SuperCollider. Trying here to find differrent ways of
>> extracting data from the sensor signals. Maybe even using Kalman
>> filtering. I've been using  Nescivi's SenseWorld, which is really  
>> nice.
>
> Thanks. Glad to hear that they are being used :)
> I am interested in any suggestions, since I am still developing them  
> further.
>
> the SenseWorld classes come from working on an interactive dance  
> piece called
> Schwelle, which we have been performing a couple of times now.
> We have written papers about it for last year's NIME and ICMC, so  
> you may want
> to check these out for some background information, and references  
> to other
> projects.
>
> Getting actual position measurements with accelero's is tricky, as  
> you lack
> two integration constants, when you integrate from acceleration to  
> speed to
> position.
> I'll take a closer look at the article you posted, it shouldn't be  
> possible to
> implement that in SC. Did you have a go at that already?
>
>
> sincerely,
> Marije
> _______________________________________________
> sc-users mailing list
> sc-users@...
> http://lists.create.ucsb.edu/mailman/listinfo/sc-users

_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[nescivi]

It's a quark since a few weeks, if not months.
so it's in svn there.

sincerely,
Marije

On Thu, May 8, 2008 at 9:01 PM, Scott Wilson <i@...> wrote:

> Marije,
>
>  are the SenseWorld classes available?
>
>  S.
>
>
>
>  On 8 May 2008, at 08:49, nescivi wrote:
>
>
> > On Wednesday 07 May 2008 07:28:50 Eirik Blekesaune wrote:
> >
> > > Hi,
> > > I'm currently working on an interactive danceperformance, using
> > > accelerometers. I'm now trying to implement some kind of position
> > > measurements with the signals from the accelerometers. During my
> > > googling I've found this article:
> > >
> http://www.gamasutra.com/view/feature/1494/wheres_the_wiimote_using_kalman_
> > > .php?page=1
> > >
> > > I wondered if someone has done anthing with processing accelerometer
> > > data using SuperCollider. Trying here to find differrent ways of
> > > extracting data from the sensor signals. Maybe even using Kalman
> > > filtering. I've been using  Nescivi's SenseWorld, which is really nice.
> > >
> >
> > Thanks. Glad to hear that they are being used :)
> > I am interested in any suggestions, since I am still developing them
> further.
> >
> > the SenseWorld classes come from working on an interactive dance piece
> called
> > Schwelle, which we have been performing a couple of times now.
> > We have written papers about it for last year's NIME and ICMC, so you may
> want
> > to check these out for some background information, and references to
> other
> > projects.
> >
> > Getting actual position measurements with accelero's is tricky, as you
> lack
> > two integration constants, when you integrate from acceleration to speed
> to
> > position.
> > I'll take a closer look at the article you posted, it shouldn't be
> possible to
> > implement that in SC. Did you have a go at that already?
> >
> >
> > sincerely,
> > Marije
> > _______________________________________________
> > sc-users mailing list
> > sc-users@...
> > http://lists.create.ucsb.edu/mailman/listinfo/sc-users
> >
>
>

_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[marierm]

>>
> Thanks. Glad to hear that they are being used :)
> I am interested in any suggestions, since I am still developing them  
> further.
>
I have been using them with accelerometers too.  I found them very  
useful.  I saved a lot of time.  Concerning suggestions, I was  
thinking about fltering capabilities.  .mean does a good low pass, but  
is there anything close to a high pass in the class ?

And Eirik, I'm sorry I cannot help you with your Kalman filtering.  I  
have only been using accelerometers as tilt sensors.

Martin Marier
_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[Eirik Arthur Blekesaune]

For some reason i have to specifically update SenseWorld in Quarks,  
when I try to update the whole Quarks library, SenseWorld is not  
updated. :-|

btw the ParametreSpace class is really cool. I've been working on  
something similar, but this implementation looks very promising. So it  
seems that once again, using SuperCollider, the best code is the one  
that I don't have to write!



-Eirik Blekesaune

Den 8. mai. 2008 kl. 21.01 skrev Scott Wilson:

> Marije,
>
> are the SenseWorld classes available?
>
> S.
>
> On 8 May 2008, at 08:49, nescivi wrote:
>
>> On Wednesday 07 May 2008 07:28:50 Eirik Blekesaune wrote:
>>> Hi,
>>> I'm currently working on an interactive danceperformance, using
>>> accelerometers. I'm now trying to implement some kind of position
>>> measurements with the signals from the accelerometers. During my
>>> googling I've found this article:
>>> http://www.gamasutra.com/view/feature/1494/wheres_the_wiimote_using_kalman_
>>> .php?page=1
>>>
>>> I wondered if someone has done anthing with processing accelerometer
>>> data using SuperCollider. Trying here to find differrent ways of
>>> extracting data from the sensor signals. Maybe even using Kalman
>>> filtering. I've been using  Nescivi's SenseWorld, which is really
>>> nice.
>>
>> Thanks. Glad to hear that they are being used :)
>> I am interested in any suggestions, since I am still developing them
>> further.
>>
>> the SenseWorld classes come from working on an interactive dance
>> piece called
>> Schwelle, which we have been performing a couple of times now.
>> We have written papers about it for last year's NIME and ICMC, so
>> you may want
>> to check these out for some background information, and references
>> to other
>> projects.
>>
>> Getting actual position measurements with accelero's is tricky, as
>> you lack
>> two integration constants, when you integrate from acceleration to
>> speed to
>> position.
>> I'll take a closer look at the article you posted, it shouldn't be
>> possible to
>> implement that in SC. Did you have a go at that already?
>>
>>
>> sincerely,
>> Marije
>> _______________________________________________
>> sc-users mailing list
>> sc-users@...
>> http://lists.create.ucsb.edu/mailman/listinfo/sc-users
>
> _______________________________________________
> sc-users mailing list
> sc-users@...
> http://lists.create.ucsb.edu/mailman/listinfo/sc-users

_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[nescivi]

Hi,

On Thursday 08 May 2008 23:28:34 marierm@... wrote:
> > Thanks. Glad to hear that they are being used :)
> > I am interested in any suggestions, since I am still developing them
> > further.
>
> I have been using them with accelerometers too.  I found them very
> useful.  I saved a lot of time.  Concerning suggestions, I was
> thinking about fltering capabilities.  .mean does a good low pass, but
> is there anything close to a high pass in the class ?

Because of the data we got in in our application (somewhat jumpy wireless
connection due to electro magnetic interference in the theater), lowpass
filtering was more necessary on the raw data.
Standard deviation is a nice way to use the amount of change in a certain time
window.

I have been looking at moving more things to the server side for processing,
and been exploring the usabilitly of various existing UGens for this. That's
not quite finished yet though.

sincerely,
Marije
_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users

[nescivi]

Hi Eirik,

On Friday 09 May 2008 04:15:34 Eirik Blekesaune wrote:
> For some reason i have to specifically update SenseWorld in Quarks,
> when I try to update the whole Quarks library, SenseWorld is not
> updated. :-|

that is weird...

> btw the ParametreSpace class is really cool. I've been working on
> something similar, but this implementation looks very promising.

Thanks.
I wrote that class for a piece I just performed about a week or so ago.
Slightly inspired on Ali Momeni's 2D parameter spaces (paper on the NIME
2003), but a different weighting algorithm for the interpolation, than he
proposed.

sincerely,
Marije
_______________________________________________
sc-users mailing list
sc-users@...
http://lists.create.ucsb.edu/mailman/listinfo/sc-users