is there's a way to automatically find peaks in an histogram?

Hello,

I have a script to analyze data from a detector (the data is not necessarily repeated). right now I have to find the peaks manually which is time-consuming!

is there's a way I can get the range of the peak base in the x-axis automatically? and is there's a way to neglect small peaks (let's say small in respect to the ratio from the max peak)?

here is an example of histogram I'm working with

[Rik]

Have you tried the findpeaks function?

[sani]

yes but I couldn't applied it for histogram, just to my original data (and it dosn't relevant)

Answer 1

[Steven Lord]

You can do this using the islocalmax function.

% Sample data
rng default
x = randn(1, 1e5);
h = histogram(x);

% Retrieve some properties from the histogram
V = h.Values;
E = h.BinEdges;

% Use islocalmax
L = islocalmax(V);

% Find the centers of the bins that islocalmax identified as peaks
left = E(L);
right = E([false L]);
center = (left + right)/2;

% Plot markers on those bins
hold on
plot(center, V(L), 'o')

[sani]

it's doing the job, but I need to determine some tolerance because my histogram is noisy and this is what I get. It recognizes every bump as a peak, which is correct, but not what I need.

[Steven Lord]

The documentation for islocalmax lists several options that you can use to customize how islocalmax looks for peaks. MinProminence, MinSeparation, and/or MaxNumExtrema may be of use depending on whether you want peaks that are sufficiently higher than their surroundings, sufficiently far apart, or only a limited number of peaks.

[Image Analyst]

How is islocalmax() different than findpeaks()? Is it just that it's in base MATLAB instead of the Signal Processing Toolbox?

[sani]

thanks a lot Steven! it was really helpful:)