How to create plots for special areas of an array?

Dear people,

I have a plot like this:

I want to plot each of the four waves separately. Of course I could just divide the whole length by 4 but then there would still be the noise from between the big waves. I tried to create a new array where all values smaller than 0.03 are deleted. But then I also delete those values of the big waves as well. Does anybody has an idea how to solve this? Thank you very much:)

[PromoCodeBSJ]

Are the plots subject to change? If not, I would just figure out the x range of the four graphs and then graph those

[Stwiwi]

Thank you for the quick response. Unfortunately the plots are created in a loop with changing data arrays in every loop. The plots from each loop look similar but are not the same, so the starting points of the big waves vary.

[PromoCodeBSJ]

Are you acquiring data from an external source?

[jonas]

You could calculate a moving std and remove the segments with the lower variation. Could you upload the data? Are there always four distinct waves?

[Stwiwi]

Yeah there are always four waves. I attached three data files. Always with four waves but with different lengths and values.

Answer 1

[Star Strider]

This is perfect for the findchangepts (link) function, introduced in R2016a.

The Code —

D = load('005a_100Hz.mat');
Ch9 = D.Channel_9_Data;
tv = D.Channel_1_Data;                                              % Time Vector

[lpt,rsd] = findchangepts(Ch9, 'Statistic','rms', 'MaxNumChanges',8);
for k1 = 1:2:numel(lpt)-1
    sig_seg{(k1+1)/2,1} = Ch9(lpt(k1):lpt(k1+1));                   % Signal Values At Each Segment
    sig_seg{(k1+1)/2,2} = lpt(k1):lpt(k1+1);                        % Index Range
end    

vl = bsxfun(@times, ones(numel(lpt), 2), [min(Ch9) max(Ch9)]);      % Create Vertical Lines For ‘figure(1)’

figure(1)
plot(tv, Ch9)
hold on
plot([tv(lpt),tv(lpt)]', vl', '-r', 'LineWidth',1.5)
hold off
grid

figure(2)
for k1 = 1:size(sig_seg,1)
    subplot(size(sig_seg,1), 1, k1)
    plot(tv(sig_seg{k1,2}), sig_seg{k1,1})
    xlabel('Time (s)')
    grid
end

The Plots —

[Stwiwi]

Thanks so much it works perfectly.

[Star Strider]

As always, my pleasure.

[Stwiwi]

Unfortunately this method wont work for the attached data file. Probably because of all the outliers? Any idea how to solve this? That would be great!

Answer 2

[Image Analyst]

What I would do is to scan the signal with a filter to detect the large variations zones, like with movmax(), imdilate(), stdfilt() or similar. Then I'd threshold to find the "quiet" low signal areas. Then I'd find the centroid of those zones, assuming you want to split the signal in the middle of the quiet zones, with the regionprops() function (in the Image Processing Toolbox). I have done this exact thing for a nearly identical signal someone had a couple of years ago. Attach your data if you want any more help.

Answer 3

[Stwiwi]

Ok I have attached one of the Data files and my code so far:

Ord              =    6;                                                
f_cf             =    5;                                                
f_s              =    100;                                              
[b_low, a_low]   =    butter(Ord,f_cf/(0.5*f_s),'low');                 

files = dir('*.mat');

 for t=1:length(files)

    load([files(t).name]);

    % Filter Data and write in Workspace
    ForceX  = filtfilt(b_low,a_low,detrend(Channel_19_Data));
    ForceY  = filtfilt(b_low,a_low,detrend(Channel_20_Data)); 
    ForceZ  = filtfilt(b_low,a_low,detrend(Channel_21_Data)*-1); 
    USS    = detrend(Channel_9_Data*-1);
    time    = (Channel_1_Data);
end

In the for loop i then go in a function which plots every test and saves it, but I think this code should be enough because I attached only one data file. When I plot USS it shows the plot from my original question. Does that help?