how to calculate duration of peaks

I have an envelope signal attached y.mat and time array t.mat

I want to calculate the duration in second between the peaks (mentionned in red color in the figure below)

How I can reach this task

Answer 1

[Star Strider]

Try this —

LD1 = load('t.mat');

ty = LD1.ty.';

LD2 = load('y.mat');

sh1 = LD2.sh1;

y = sh1.';

[pks,plocs] = findpeaks(y, 'MinPeakProminence',0.5); % Desired Peak & Index

[vys,vlocs] = findpeaks(-y, 'MinPeakProminence',0.0000001); % Valleys & Indices

for k = 1:numel(pks)

vlcs(1,k) = max(vlocs(vlocs<=plocs(k))); % First Valley Index (Beginning Of Spike)

vlcs(2,k) = min(vlocs(vlocs>=plocs(k))); % Last Valley Index (End Of Spike)

end

deltat = -diff(ty(vlcs));

figure

plot(ty,sh1)

hold on

plot(ty(plocs), pks, '^r')

plot(ty(vlcs), sh1(vlcs),'.r')

hold off

grid

xlabel('Time')

ylabel('Amplitude')

text(mean(ty(vlcs)),zeros(size(plocs)), compose('\\leftarrow \\Delta t = %.6f',deltat), 'Horiz','left', 'Vert','middle', 'Rotation',60)

producing —

EDIT — (9 Jul 2022 at 17:24)

Minor edit to my original code to make it more robust and more efficient. It produces the same reault. I have no idea how to make it more robust with respect to other data sets, since I have only one data set to work with here, and it works correctly. It will be necessary to experiment with it to get it to work with other data sets. One possibiity is to experiment with the name-value pair arguments in the second findpeaks call to be certain that it detects the beginning and end ‘valleys’.for each peak, and rejects any noise that may be in the data. My code requires that each peak has valleys on either side of it, since they are in the posted data.

EDIT — (9 Jul 2022 at 22:58)

This slightly revised code works for both sets of files —

% LD1 = load('Abdelhakim Souidi t.mat');

LD1 = load('Abdelhakim Souidi (2) t.mat');

ty = LD1.ty.';

% LD2 = load('Abdelhakim Souidi y.mat');

LD2 = load('Abdelhakim Souidi(2) y.mat');

sh1 = LD2.sh1;

y = sh1.';

[pks,plocs] = findpeaks(y, 'MinPeakProminence',0.1, 'MinPeakDistance',250); % Desired Peak & Index

[vys,vlocs] = findpeaks(-y, 'MinPeakProminence',0.0001, 'MinPeakDistance',500); % Valleys & Indices

for k = 1:numel(pks)

vlcs(1,k) = max(vlocs(vlocs<=plocs(k))); % First Valley Index

vlcs(2,k) = min(vlocs(vlocs>=plocs(k))); % Last Valley Index

end

deltat = diff(ty(vlcs));

figure

plot(ty,sh1)

hold on

plot(ty(plocs), pks, '^r')

plot(ty(vlcs), sh1(vlcs),'.r')

hold off

grid

xlabel('Time')

ylabel('Amplitude')

text(mean(ty(vlcs)),zeros(size(plocs)), compose('\\leftarrow \\Delta t = %.6f',deltat), 'Horiz','left', 'Vert','middle', 'Rotation',60)

producing —

It matters not which peak of the double-peak envelopes are identified as the actual peak, since the baseline duration is the desired result, and that simply depends on the relation of individual ‘plocs’ indices and the nearest ‘vlocs’ indices.

.

[Lam Ha]

I follow your code and apply it on my data but I got the problems like it. It marks the higest point with x=y. I'm looking forward to hear suggestion from u.

This is my data

[Star Strider]

My code does exactly what it is supposed to do with your data.

Please post this as a new Question, specifically with a description of what you want to do, and include the URL of this thread.

I will not respond to it further here, since it is not directly relevant to this thread.

Answer 2

[Image Analyst]

Try this:

clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;  % Erase all existing variables. Or clearvars if you want.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 15;
st = load('t.mat')
t = st.ty;
sy = load('y.mat')
sh1 = sy.sh1;
y = sh1;
plot(t, y, 'b-', 'LineWidth', 2);
grid on;
hold on;
xlabel('t', 'FontSize', fontSize);
ylabel('y', 'FontSize', fontSize);
[peakValues, indexesOfPeaks] = findpeaks(y, 'MinPeakProminence', 0.25); % Desired Peak & Index
% Plot peaks
plot(t(indexesOfPeaks), peakValues, 'rv')
numPeaks = numel(peakValues)
% Fall down the left sides
leftIndexes = zeros(numPeaks, 1);
tLeft = zeros(numPeaks, 1);
for k = 1 : numPeaks
    for k2 = indexesOfPeaks(k) : -1 : 1
        thisValue = y(k2);
        if thisValue < 0.1 && thisValue > y(k2+1)
            leftIndexes(k) = k2;
            tLeft(k) = t(leftIndexes(k));
            % Put a green line at the start of the peak.
            xline(tLeft(k), 'Color', 'g', 'LineWidth', 2);
            break;
        end
    end
end
% Fall down the right sides
rightIndexes = zeros(numPeaks, 1);
tRight = zeros(numPeaks, 1);
for k = 1 : numPeaks
    for k2 = indexesOfPeaks(k) : length(y)
        thisValue = y(k2);
        if thisValue < 0.1 && thisValue > y(k2-1)
            rightIndexes(k) = k2;
            tRight(k) = t(rightIndexes(k));
            % Put a red line at the end of the peak.
            xline(tRight(k), 'Color', 'r', 'LineWidth', 2);
            break;
        end
    end
end
% Get the widths
peakWidths = tRight - tLeft

The left side is tLeft and is indicated by the green line.

The right side is tRight and is indicated by the red line.

The widths are given by peakWidths.

peakWidths =

0.146280579131303

0.157264103844234

0.155516724912631

0.159510733899151

0.170993509735397

0.159011482775836

0.177983025461807

0.179980029955067

0.150773839241138

0.159760359460809

[Abdelhakim Souidi]

@Image Analyst you are the best. If you don't mind please how can I implement findchgpts to detect the boundaries

thank you

[Image Analyst]

@Abdelhakim Souidi I don't use it that much. All I'd do is look at the documentation and the examples there and try to apply it to your data, something you can do as well as me.

Answer 3

[Image Analyst]

Try findchangepts or islocalmin

[Abdelhakim Souidi]

@Image Analyst I try it with findchangepts and islocalmin however with findchangepts it did not detect it very well

With islocalmin there are other few indesirable points detected how can I smooth it

[Abdelhakim Souidi]

Can you please try it on the data I provided

[Image Analyst]

Looks like Star's code worked, so just use that.

[Abdelhakim Souidi]

@Image Analyst I tried the @Star Strider 's code unfortunately it doesn't work on all the envelopes I have

[Image Analyst]

Please give an example of someplace in your data where Star's code identified either the start of the peak, the peak location, or the end of the peak as being at the wrong place.

Another option would be to just find the peaks with findpeaks and then "fall down" each side of the peak until you get to a point where the data starts to increase again.

[Abdelhakim Souidi]

@Image Analyst here is an example

[Image Analyst]

Try adjusting the 'MinPeakDistance' to make sure your valleys are not too close together.