Hi Popesz,

At least on kinect, there's always a bit of noise in the tracking.
The change from frame-to-frame will never be exactly 0, so you need
something that acts more like a threshold.

One place to start would be collecting some data on the amplitude of
the noise when you're doing what you'd consider standing still.
For one dimensional data, you just want to see how big is the typical
value of |x(n)-x(n-1)|.  For more dimensions, you need sqrt sum of
squares on the differences

Then, I'd probably give it a exponential decay weighting and compare
against a threshold that is just some factor larger than the basic
noise amplitude.

1-D:
[r coordinate]
|
[t f f]
X
[- 0]
|   [r fb]
|   [* 0.9] or some other factor <1
[+ 0]
|\
| [s fb]
|
[expr $f1 < 10]
|
[select 0 1]

My kinect data is all over the place.  The device is really sensitive
to lighting, background detail, and so on.  You should consider that
the noise may be larger sometimes depending on where you're using the
sensor.

Chuck


On Sat, Feb 1, 2020 at 2:58 PM Csaba Láng <langcsaba@gmail.com> wrote:
>
> Dearest list,
>
> is there any object which only gives me a value from a constantly changing random numbers when it stops for a while, let's say for one second or so?
> I am looking for the parameters of tracking hands from Kinect like motion detector when the hands stops on an object for a while.
>
> Best,
>
> Popesz
> _______________________________________________
> Pd-list@lists.iem.at mailing list
> UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list