[Scilab-users] "Smoothing" very localised discontinuities in curves.

[Stéphane Mottelet]

Why not use median filtering with a small window ? 

> Le 4 avr. 2016 à 07:57, tim at wescottdesign.com a écrit :
> 
> As written (with or without the commenting out) that code won't produce phase shifts.  In technical terms it's a FIR filter that is symmetrical around zero delay, and such filters do not introduce phase shifts.
> 
>  
> 
>> On 2016-04-03 22:47, Claus Futtrup wrote:
>> 
>> Hi Buk
>> 
>> When data goes zig-zag like that, I'm sometimes successful with a Bartlett (triangular) smoothing window of only very few points (like 3 points).
>> 
>> // function [out]=bartlett3p(indata) // for smoothing (when plotting)
>>     // out = indata;
>>     // if length(indata)>2 then
>>         // out(2:$-1) = 0.25*indata(1:$-2) + 0.5*indata(2:$-1) + 0.25*indata(3:$);
>>     // end // Bartlett Window = Triangular window
>> endfunction
>> 
>> Yes, it will affect your data, probably introduce phase shifts.
>> 
>> Best regards,
>> Claus
>> 
>>> On 03-04-2016 21:09, scilab.20.browseruk at xoxy.net wrote:
>>> HI,
>>> 
>>> The data I'm dealing with is experimentally produced; and thus contains occasional, localised discontinuities (inflections), that I need to remove before that data is suitable for is use in FEM modeling software, which requires that it be strictly monotonic. The attachment shows the full curve plus a close up of a couple of examples of the type of discontinuity I need to deal with.
>>> 
>>> I haven't yet decided whether to simply omit points (thus connect A to F & G to J) or whether to retain the same number of points by interpolating new points onto that line as shown in red.
>>> 
>>> I've looked and played several of the smoothing, convolution and interpolation routines that scilab provides, but (besides that I don't understand the output some of them produce) they also seem to affect the data more than I would like. Some seem to introduce a 'phase shift'; others smooth out larger scale bumps in the curve that need to be retained; and others generate many extra points which I don't think is helpful, the FEM software is going to do its own interpolations anyway.
>>> 
>>> 
>>> But the bit I'm asking about here is how to detect point A&F and G&J? 
>>> 
>>> Any thoughts or pointers as to a) the algorithm to use; b) how to implement it in SciLab?
>>> 
>>> Cheers, Buk.
>>> 
>>> ____________________________________________________________
>>> Can't remember your password? Do you need a strong and secure password?
>>> Use Password manager! It stores your passwords & protects your account.
>>> Check it out at http://mysecurelogon.com/password-manager
>>> 
>>> 
>>> _______________________________________________
>>> users mailing list
>>> users at lists.scilab.org
>>> http://lists.scilab.org/mailman/listinfo/users
>> 
>> 
>> _______________________________________________
>> users mailing list
>> users at lists.scilab.org
>> http://lists.scilab.org/mailman/listinfo/users
>  
> 
>  
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20160404/faafa672/attachment.htm>