seperating higher and lower levels of a pulse

I have data which contains pulses with large number of ripples. How do i identify the maximum(mean) and minimum(mean) of each pulse. Also the data varies in pulsewidth in other data i have(not attached).

I searched and got statelevels which could be used , but i do not have the signal processing toolbox and need to find it mathematically and accurately.

Answer 1

[Image Analyst]

Please post a screenshot (a PNG file - it's easier for us to see immediately without saving your fig and then moving over to MATLAB to open it and then view it.)

Not having see it, I guess I'd first suggest thresholding. If the signal is above some value, it's the top of the pulse and if it's below the value, it's the bottom of the pulse train.

threshold = mean([min(pulseTrain), max(pulseTrain)]); % Half way between min and max.
pulseTopIndexes = pulseTrain > threshold;
% Get the overall min and max of the pulse tops.
maxPulseTop = max(pulseTrain(pulseTopIndexes));
minPulseTop = min(pulseTrain(pulseTopIndexes));

If you need the min and max of each individual pulse separately, then you'll have to label each pulse and examine it. The easiest way to do that is to use the Image Processing Toolbox (untested code)

props = regionprops(pulseTopIndexes, 'PixelValues');
for k = 1 : length(props)
    maxPulseTop(k) = max(props(k).PixelValues);
    minPulseTop(k) = min(props(k).PixelValues);
end

[SHANTANU KSHIRSAGAR]

I have attached herewith the image of data i am dealing with

[Image Analyst]

Looks like my code should work. Attach your actual data in a .mat file if you still have trouble.

[SHANTANU KSHIRSAGAR]

The min and max value of pulse could be found using the first method. i could not make use of the 2nd method for finding it for all pulses. The use of the pulse and transition matrix for the same(to find out the state levels and then set a references that we need to calculate ) is possible, but due to unavailaility of the Signal processing toolbox some generalisation is required.

Something equivalent to the method of finding state levels and setting references can be done, in a generalized manner(mathematical way) .

[Image Analyst]

I suggest you histogram the data and fit the two humps in the histogram to Gaussians with the attached fit_two_Gaussians demo.

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;

fileName = 'data.xlsx';
[numbers, strings, raw] = xlsread(fileName);
x = numbers(:, 1);
y = numbers(:, 2);
subplot(2, 2, 1);
plot(x, y, 'b-');
grid on;
title(fileName, 'FontSize', fontSize, 'Interpreter', 'none');
xlabel('x', 'FontSize', fontSize);
ylabel('y', 'FontSize', fontSize);
% Histogram the data.
subplot(2, 2, 2);
histogram(y);
grid on;
caption = sprintf('Histogram of %s', fileName);
title(caption, 'FontSize', fontSize, 'Interpreter', 'none');
xlabel('y', 'FontSize', fontSize);
ylabel('Count', 'FontSize', fontSize);

fileName = 'data_1.xlsx';
[numbers, strings, raw] = xlsread(fileName);
x1 = numbers(:, 1);
y1 = numbers(:, 2);
subplot(2, 2, 3);
plot(x1, y1, 'b-');
grid on;
title(fileName, 'FontSize', fontSize, 'Interpreter', 'none');
xlabel('x1', 'FontSize', fontSize);
ylabel('y1', 'FontSize', fontSize);
% Histogram the data.
subplot(2, 2, 4);
histogram(y);
grid on;
caption = sprintf('Histogram of %s', fileName);
title(caption, 'FontSize', fontSize, 'Interpreter', 'none');
xlabel('y1', 'FontSize', fontSize);
ylabel('Count', 'FontSize', fontSize);

[SHANTANU KSHIRSAGAR]

Thanks Image analyst this is helpful. But can somthing be used instead of the ''fitnlm'' command as i have just the basic matlab and image processing toolbox with me, no other tool box is available with me. And also i do not have to change the raw data as it will lose its originality during calculation.

keeping the Signal near to original i need to calculate the min and max of each pulse and in anautomated manner. That is why i mentioned ''The use of the pulse and transition matrix for the same(to find out the state levels and then set a references that we need to calculate ) is possible'' in previous comment. Can we do anything in parallel to it with basic MATLAB toolboxes.

[Image Analyst]

You might try findchangepts() but you can see that where the pulse starts is not exactly a definitely place. Just look at the histogram. How about taking the max of the signal and just thresholding it at that minus 0.05? Then for each pulse, keep going from the starting point until the signal begins to turn around, and take the stats from there on.

[SHANTANU KSHIRSAGAR]

I am trying the same, will be asking you if furthur doubt situation arrises during execution. Thank you supportiveness and corresponding solutions.