[PD-dev] how to contribute

[cyrille henry]

hello,

This abstraction should work like your external.
So you can easilly use it on every platform...

But after more reflexion, i realize that the noise immunity is not very smart, since it is not constant over regarding the offset of the signal.
i.e : if f a time variant function and k a constant, local_max(f) + k != local_max(f+k).

i would prefer the use of a constant threshold better than a threshold depending on the amplitude of the signal.

So, i don't think i'll include this abstraction on the mapping directory.
unless someone have good arguments...

doing the same kind of analysis but with a constant threshold is just like using 2 mapping object : 
hysteresis and local_max. 
it's a pre-filtering technics, but with a "strange" filter.
hysteresis is a very good noise reduction filter for noisy data...
(i should develop more it's helpfile)

Cyrille





Loic Kessous a écrit :
> Hi All,
> 
> I have an external that I made and use a lot (in fact I have several  
> but this one is a priority for me now). I compiled it for os X, but I  
> would be happy to use it also for MSW and even Linux.
> I will be glad to let others use it too, I talked with Cyril Henry and  
> suggest me to add it to pd-extended so it will be compiled on the farm  
> for all the platforms and included in the distribution.
> This external does the same thing as Cyril's [local_max] abstraction  
> but use a different method for robustness to noisy signals and avoid  
> inconvenient that can be introduce by pre-filtering.
>   As it was easier to do it as an external than an abstraction, that  
> what I did.
> 
> So can someone tell me how if it is possible and what I have to do for  
> this.
> 
> loic
> ---
> 
> _______________________________________________
> Pd-dev mailing list
> Pd-dev at iem.at
> http://lists.puredata.info/listinfo/pd-dev
> 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: maxloc.pd
URL: <http://lists.puredata.info/pipermail/pd-dev/attachments/20090118/da21e31e/attachment.txt>