MinPeakProminence error in findpeaks - detecting unwanted signal

I am trying to use the findpeaks function to look for sharp and intense peak in 100 spectra. Here is an example of my spectrum:

I tried different name-value arguments so that it would look for sharp and intense peaks only. However, it seems like the 'MinPeakProminence' is not working as expected.

When I apply the following line to the spectrum above (I limited the number of peaks (NPeaks) to 1 for simplier visualization) :

load ('Data.mat'); figure;
findpeaks(y, 'MinPeakProminence',0.15,'NPeaks',1,'MaxPeakWidth',20,'MinPeakDistance',7,'Annotate','extents');

This is what it gives me:

It seems like the function is drawing its own borders and using the boaders as lowest left and right intervals (This is just my observation, not sure if I intepreted that correctly).

Below shows the magnified detected "peak":

Has anyone encounter the same issue? Any help would be greatly appreciated.

Thanks in advance!

[Mathieu NOE]

and what is / are the peaks you are interested in ?

[See Yoong Wong]

@Mathieu NOE

Thanks for your reply

I am interested in the prominent and sharp peaks. I have hundreds of different spectra like this and trying to use matlab to scan through the spectra.

For this particular spectrum, since I set the minimum prominence value at 0.15, findpeaks should not detect any peak but it did. And the peak it detected has an actual prominence value lower than 0.05.

[Mathieu NOE]

hello

I understand you don't get what you want with findpeaks , but that still does not tell me what you want (a picture is better that text ....)

[Mathieu NOE]

in other words , maybe there is better alternative (if I get what you're looking for)

[See Yoong Wong]

@Mathieu NOE I see! I am looking for peaks like this one:

Peaks like this one might appear anywhere along the x-axis. They can have varied intensity (prominence between 0.1 to 0.95), but they should be sharp (width between 3 to 15 datapoints).

Answer 1

[Mathieu NOE]

hello again

this would be my suggestion

hope it helps

load ('Data.mat'); 
[PKS,LOCS,W,P] = findpeaks(y, 'MinPeakProminence',0.15,'MaxPeakWidth',15,'MinPeakDistance',100);
% rejet peaks above 0.9 (the ones that could be on the y = 1 plateau)
id = (PKS<0.9);
PKS = PKS(id);
LOCS = LOCS(id);
W = W(id);
P = P(id);
% then select the first one (along the x axis)
PKS = PKS(1);
LOCS = LOCS(1);
W = W(1);
P = P(1);
figure;
plot(y); hold on 
plot(LOCS,PKS,'dr');

[See Yoong Wong]

Thank you very much for your help Mathieu!

[Mathieu NOE]

as always, my pleasure !