Re: findpeak -> peakdet.m ?

[Olivier Crouzet]

Hi,

I Don't know if it's relevant (perhaps you really want to program it
yourself, perhaps you know it and it doesn't exactly fit your needs) but
Eli Billauer provides a script named peakdet.m
(http://billauer.co.il/peakdet.html) which does exactly that: locate
peaks (and valleys too) and returns 'y' and, either 'x' or the vector
indices. It also allows to specify a delta when one wants only global
peaks, not every local one (particularly usefull when there is noise
in the data). I often use it these times and it does a very good job
(and it's also available for python and C; I haven't tested these but
links are provided on the web page).

The bottom side is that it uses a "for" loop and may therefore take a
long time on huge data, but I suspect there's perhaps no other way when
one wants to parameterize a "delta" (I didn't take the time to try such
a modification)...

The script is written with matlab "style" but works as a charm within
octave.

Olivier.


On Wed,
28 Mar 2012 14:26:06 +0800 (SGT) asha g <address@hidden> wrote:

> 
> 
>  
> 
> Regardless, here's a solution that *won't work for every dataset*,
> but I think will work in your case.
> 
> % first derivative (kind of)
> dx = diff(vvl);
> % second derivative (kind of)
> dxx = diff([0,dx]);
> % peaks (peaks = -2, troughs = 2)
> idx = find(dxx == -2);
> 
> % now get the peak vvl values & corresponding x values
> vvl(idx)
> x(idx)
> 
> I tested this with a very simple dataset and it worked, but can't
> guarantee it will always work.  In particular, if the peak is
> something like this: [1,2,3,3,3,2,1], this code will not identify
> it.  Looking at the plot you sent me, it should be ok.
> 
> Andrew
> 
> Sorry, I can't get this to work but is there something in octave
> similar to Matlab's pks = findpeaks(data)
> [pks,locs]
> = findpeaks(data) returns the
> indices of the local peaks.
> 
> If I can get this, then my problem is solved.
> Thanks
> Asha 


-- 
  Olivier Crouzet, PhD
  Laboratoire de Linguistique -- EA3827
  Département de Sciences du Langage
  UFR Lettres et Langages
  Université de Nantes
  Chemin de la Censive du Tertre - BP 81227
  44312 Nantes cedex 3
  France

     phone:        (+33) 02 40 14 14 05 (lab.)
                   (+33) 02 40 14 14 36 (office)
     fax:          (+33) 02 40 14 13 27
     e-mail:       address@hidden
                
  http://www.lling.fr/