Finding peak value and peak arrival time from a time-domain plot and store it into a matrix.

I have 360 differential signals to be processed through SVM. How do I extract peak value and peak arrival time from time-domain plot of each signal and store it in a single matrix? Below is my code.

 % load a .mat file
   load ('Defect_Signals_with_Labels.mat');
   fs  = 100000;

   for j = 1:360

       x = SurfaceDifferential(:,j);

       % set time to be in milliseconds
       time = ( 0: length(x)-1) / fs;

       % plot frequency in time domain
       figure(2);
       plot(time, x);
       xlabel('Time (s)');
       ylabel('Magnitude (Gauss)');
       grid on;

   end

Answer 1

[Sandro Lecci]

Hi Syed,

are you looking for the max value and its index ?

peakValues = NaN(360, 2); %column 1 is peak value, column 2 is arrival time

   for j = 1:360
       % Extract the time-series data
       x = SurfaceDifferential(:,j);

       % set time to be in milliseconds
       time = ( 0: length(x)-1) / fs;

       %extract max and its location from the timeseries data
       [xMax, xIdx] = max(x);
       peakValues(j,1) = xMax;
       peakValues(j,2) = time(xIdx); %to get the value in milliseconds

       %plot...
   end

best, Sandro

[Syed Farez]

Hai Sir,

Yes. I wanted to get the max value for the frequency. And for time, I wanted to get the rise time if possible. Sorry for the misleading title. It should be the rise time.

[Sandro Lecci]

Hello,

from my personal experience, getting the correct rise time is not as trivial as identifying the max value (with a sample to sample resolution, of course). As I have no clue about how your data look like I can simply propose you an approach.

Assuming the event you are interested in stands out clearly from the background noise, I would calculate the first derivative of your signal and identify the rising moment.

Try to calculate diff(x) and see by eye ( figure(); plot(diff(x)) ) whether the result is satisfactory enough to clearly identify the correct rise time.

Best, Sandro

[Guillaume]

Sandro makes some good points in his latest comment.

I just want to point out that the loop in the answer is completely unnecessary and the whole lot can be replaced by just:

[peakvalues, where] = max(SurfaceDifferential);
correspondingtime = time(where);

[Syed Farez]

That's a very good and helpful information for me. I think the max value will give a better result for my work. Thank you very much for this Sandro & Guillaume. Much appreciated indeed.