How do find the peaks of pulses from an array

I have an array that is about 1million x 2 I need help finding the peaks of each pulse of the entire data. Any ideas?

The data set is here below could any one give me an example of getting a peak for each Pulse throughout the entire data set?

Answer 1

[Star Strider]

It would be helpful to have a sample of the data.

Two options: findpeaks and islocalmax.

[Mama Seck]

this is the graph I will upload the m.file in one minute

[Mama Seck]

this is the m file youll need the mat file to open it but you can only upload upto 5MB here

[Star Strider]

The .m file isn’t at all helpful because the data don’t exist.

That aside, the image is helpful. If you want to identify clusters of peaks, use the envelope function to create the upper (and lower, if necessary) envelopes, then use findpeaks or islocalmax to identify the peaks in the envelope instead of the peaks of the original signal, for each signal.

[Mama Seck]

I've talored the data to someting that can be uploaded here. The can you give me an example of this, so that I can apply it to my data set? plead and thank you.

[Star Strider]

I am not certain what result you want, and I still cannot make any sense out of the .m file.

Try this —

LD = load('Sig_2_Bucket_Flicking_1_106_Dogbone_9mm_5_6_2021_Xaxis_Tailored_Data2.mat');
data_seg = LD.data_seg;

[ds_envu,ds_envl] = envelope(data_seg(:,2), 150, 'peak');
[pks,locs] = findpeaks(ds_envu, 'MinPeakProminence',10);

figure
plot(data_seg(:,1), data_seg(:,2))
hold on
plot(data_seg(:,1), ds_envu, '-r')
plot(data_seg(locs,1), ds_envu(locs), '^g', 'MarkerFaceColor','g')
hold off
grid
% xlim([3.5 4.15])

The envelope call appears to do cluster the peaks correctly (at least as I view them), so that likely does not need to be changed significantly.

With respect to the findpeaks call, adjust the 'MinPeakProminence' value (here 10) or add other arguments to 'MinPeakProminence' (or replace it perhaps with 'MinPeakHeight') as necessary to get the result you want. This code detects 36 separate peaks in the data provided.

.

[Mama Seck]

Thank you so much, this is what I was going for. Can I apply a if else statement to specify the maximum or min peak in each pulse because some of these pulses the largest pulse is a negative value

.

[Star Strider]

My pleasure!

To detect negative pulses, first use ‘ds_envl’ to get the lower envelope, then use findpeaks with the negative of that to get the lower peaks.

For example —

LD = load('Sig_2_Bucket_Flicking_1_106_Dogbone_9mm_5_6_2021_Xaxis_Tailored_Data2.mat');
data_seg = LD.data_seg;

[ds_envu,ds_envl] = envelope(data_seg(:,2), 150, 'peak');
[pksu,locsu] = findpeaks(ds_envu, 'MinPeakProminence',10);
[pksl,locsl] = findpeaks(-ds_envl, 'MinPeakProminence',10);

figure
plot(data_seg(:,1), data_seg(:,2))
hold on
plot(data_seg(:,1), ds_envu, '-r')
plot(data_seg(locsu,1), ds_envu(locsu), '^g', 'MarkerFaceColor','g')
plot(data_seg(:,1), ds_envl, '-m')
plot(data_seg(locsl,1), ds_envl(locsl), 'vc', 'MarkerFaceColor','c')
hold off
grid
legend('Data','Upper Envelope','Upper Envelope Peaks','Lower Envelope','Lower Envelope Peaks', 'Location','best')
ylim([min(ylim) 500])

The ohnly way I can think of to do that is to compare the ‘locsu’ and ‘locsl’ data to locate the nearest index values and then compare the associated peak values. I attempted to come up with a straightforward way to do that, and could not.

[Mama Seck]

Awsome! this helped alot

[Star Strider]

As always, my pleasure!

I thought about one way of ‘clustering’ it, that being to threshold either the upper peak of the envelope or both the upper and lower peaks.

Try this —

LD = load('Sig_2_Bucket_Flicking_1_106_Dogbone_9mm_5_6_2021_Xaxis_Tailored_Data2.mat');
data_seg = LD.data_seg;

[ds_envu,ds_envl] = envelope(data_seg(:,2), 150, 'peak');
[pksu,locsu] = findpeaks(ds_envu, 'MinPeakProminence',10);
[pksl,locsl] = findpeaks(-ds_envl, 'MinPeakProminence',10);

ix = ds_envu > 20;                                                              % Envelope Above Threshold
dix1 = strfind(ix', [0 1]);                                                     % Leading Edge
dix0 = [strfind(ix', [1 0]) numel(ix)];                                         % Trailing Edge
EdgMtx = [dix1; dix0];                                                          % Matrix Of Leading & Trailing Edges

figure
plot(data_seg(:,1), data_seg(:,2))
hold on
plot(data_seg(:,1), ds_envu, '-r')
plot(data_seg(locsu,1), ds_envu(locsu), '^g', 'MarkerFaceColor','g')
plot(data_seg(:,1), ds_envl, '-m')
plot(data_seg(locsl,1), ds_envl(locsl), 'vc', 'MarkerFaceColor','c')
for k = 1:size(EdgMtx,2)
    idxrng = EdgMtx(1,k):EdgMtx(2,k);
    text(data_seg(fix(median(idxrng)),1), ds_envu(fix(median(idxrng))), sprintf('Max = %6.1f, Min = %6.1f',max(ds_envu(fix(median(idxrng)))), min(ds_envl(fix(median(idxrng))))), 'Vert','middle', 'Horiz','left', 'Rotation',90) 
end
hold off
grid
legend('Data','Upper Envelope','Upper Envelope Peaks','Lower Envelope','Lower Envelope Peaks', 'Location','best')
Fg = gcf;
pos = Fg.Position;
Fg.Position = pos+[-200 -500 800 500];

It defines the peak clusters, determins the maximum and minimum of each, and returns both. That shoul do essentially everything necessary, except for a few tweaks to deliver the desired information in the desired format.

EDIT — (21 May 2021 at 21:15)

Added plot figure —

.