Find peaks and valley of sinusoidal curve

Hello, I have a code that reads a file and converts it into a graph which, in turn, applies the detrend filter. However, it is not enough and I needed to know the peaks and valleys of this new graph and then subtract the values from the peaks to the valleys in order to normalize the curves. I tried to find it with findpeaks but it didn't work. How can I do it?

TY

 function startupFcn(app)
            data=readmatrix('ensaios_controlo.xlsx','sheet','analise');
             time_control=data(:,2);
             strain_control=data(:,3);
             
             t_control=seconds(time_control);
             t_control.Format = 'hh:mm:ss';

             plot(app.UIAxes,t_control,strain_control)

             s=rms(strain_control);
             app.RMS_controlEditField.Value=s;
        end

        % Button pushed function: FileButton
        function FileButtonPushed(app, event)
             [name,path]=uigetfile('.xlsx','Select a file');
             app.file=fullfile(path,name);
             data_2=readmatrix(app.file,"Sheet",'Untitled');
             time=data_2(:,2);
             strain=data_2(:,3);
                          

             x=seconds(time);
             x.Format = 'hh:mm:ss';

             d=rms(strain);
             
             app.RMSEditField.Value=d;
             
             if (1342<d)&&(d<1483)

                msgbox("Don't need analyse.","non-modal");
             
             else 
                app.TabGroup.SelectedTab=app.DataTab;
                plot(app.UIAxes3,x,strain,"r");
                

                 yline(app.UIAxes2,0)
                 yline(app.UIAxes2,-2800);
             end
        end

        % Button pushed function: AnalyseButton
        function AnalyseButtonPushed(app, event)
             data_2=readmatrix(app.file,"Sheet",'Untitled');
             time=data_2(:,2);
             strain=data_2(:,3);
             strain_2 = detrend(strain,6);
                         
             x=seconds(time);
             x.Format = 'hh:mm:ss';

             d=rms(strain);
             
             app.RMSEditField.Value=d;
             
             pks = findpeaks(strain);
             findpeaks(strain);

             pks_v= findpeaks(-strain);
             findpeaks(-strain);
             
             if (1342<d)&&(d<1483)

                msgbox("Don't need analyse.","non-modal");
             
             else 
                app.TabGroup.SelectedTab=app.AnalyseTab;
                plot(app.UIAxes2,x,strain,"r");
                hold(app.UIAxes2,'on');
                plot(app.UIAxes2,x,strain_2,"b");

                 yline(app.UIAxes2,0)
                 yline(app.UIAxes2,-2800); 
             end
        end
       

[Sam Chak]

Can you attach the "ensaios_controlo.xlsx" file?

[Mathieu NOE]

hello

FYI, if your goal is to normalize your data , you can directly do it with the matlab function

Normalize data - MATLAB normalize - MathWorks France

[Image Analyst]

If you have any more questions, then attach your data ('ensaios_controlo.xlsx') and .mlapp file with the paperclip icon after you read this:

TUTORIAL: How to ask a question (on Answers) and get a fast answer

[Maria Inês]

I apologize, I have now attached the files. The file to normalize the values that will have to be chosen in the "file" button is "test_15_seco". In the Data tab, the graph corresponding to the values in the file appears and I want to transform the maximum value into zero.

[dpb]

app.file='https://www.mathworks.com/matlabcentral/answers/uploaded_files/1353474/provete_15_seco.xlsx';

data_2=readmatrix(app.file,"Sheet",'Untitled');
time=data_2(:,2);
strain=data_2(:,3);
strain_2 = detrend(strain,6);
         
x=seconds(time);
x.Format = 'hh:mm:ss';
d=rms(strain);

app.RMSEditField.Value=d;

pks = findpeaks(strain);
findpeaks(strain);

pks_v= findpeaks(-strain);
findpeaks(-strain);

if (1342<d)&&(d<1483)
    msgbox("Don't need analyse.","non-modal");
else 
    %app.TabGroup.SelectedTab=app.AnalyseTab;
    plot(x,strain,"r");
    hold on
    plot(x,strain_2,"b");
    yline(0)
    yline(-2800); 
end

What are we to make of that???

Now explain what, specifically, you're trying to do here. It would certainly help to supply some context to the problem so folks have at least a half-chance of understanding the rationale behind the machinations--generally that is a help in devising a solution; to know the end goal and why.

Answer 1

[Mathieu NOE]

hello @Maria Inês

So far I understand you want the strain plot have all top peaks being on y = 0 line , like in your reference data file (ensaios_controlo.xlsx)

so this is my suggestion, and no I am not using findpeaks this time... I prefer peakseek which is simpler and faster and does the job very well here ( a no brainer almost)

you can access at the FEX page here ,

PeakSeek - File Exchange - MATLAB Central (mathworks.com)

or simply use the code provided below

also , no need to do any preliminary detrending, you can skip that.

Result is the new magenta line , see it has all top peaks aligned with y = 0

Zoom :

Code :

data_2=readmatrix("provete_15_seco.xlsx","Sheet",'Untitled');
time=data_2(:,2);
strain=data_2(:,3);
         
x=seconds(time);
x.Format = 'hh:mm:ss';
d=rms(strain);
app.RMSEditField.Value=d;

if (1342<d)&&(d<1483)
    msgbox("Don't need analyse.","non-modal");
else 
    %app.TabGroup.SelectedTab=app.AnalyseTab;
%     plot(x,strain,"r");
%     hold on
    plot(x,strain,"b");
    yline(0)
    yline(-2800); 
    hold on
    
    % shift strain so that it positive peaks are on the y = 0 line
    [locs, pks]=peakseek(strain);
    x_peaks = x(locs);
    yshift = interp1(x_peaks,pks,x);
    plot(x,yshift,"r");
    strain_3 = strain - yshift;
    plot(x,strain_3,"m");


end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [locs, pks]=peakseek(x,minpeakdist,minpeakh)
% FEX : https://fr.mathworks.com/matlabcentral/fileexchange/26581-peakseek?s_tid=srchtitle
% x is a vector input (generally a timecourse)
% minpeakdist is the minimum desired distance between peaks (optional, defaults to 1)
% minpeakh is the minimum height of a peak (optional)
%
% (c) 2010
% Peter O'Connor
% peter<dot>ed<dot>oconnor .AT. gmail<dot>com
if size(x,2)==1, x=x'; end
% Find all maxima and ties
locs=find(x(2:end-1)>=x(1:end-2) & x(2:end-1)>=x(3:end))+1;
if nargin<2, minpeakdist=1; end % If no minpeakdist specified, default to 1.
if nargin>2 % If there's a minpeakheight
    locs(x(locs)<=minpeakh)=[];
end
if minpeakdist>1
    while 1
        del=diff(locs)<minpeakdist;
        if ~any(del), break; end
        pks=x(locs);
        [garb, mins]=min([pks(del) ; pks([false del])]); %#ok<ASGLU>
        deln=find(del);
        deln=[deln(mins==1) deln(mins==2)+1];
        locs(deln)=[];
    end
end
if nargout>1
    pks=x(locs);
end
end

[Maria Inês]

Thanks ! Exactly what I wanted, thanks again :)

[Mathieu NOE]

My pleasure !

[Maria Inês]

Hi Mathieu,

I apologize for bothering you again but now I needed to pass the field test data through this filter and I am not getting the same results as above. Can you help me?

In figure, the filtered curve it is the blue curve and the data elminated by filter its a red.

If you need something, please tell me .

[Mathieu NOE]

hello Maria

welcome back

no problem, I can have a look but I need your new data file

[Maria Inês]

Hi,

Thank you a lot! I want to filter the data from column 2 and 3 from excel file.

I change some code to this new data.

If you need anything, just say so.

Thanks again.

Code:

[name,path]=uigetfile('.xlsx','Select a file');
             file=fullfile(path,name);
             data_2=readmatrix(file,"Sheet","Untitled");
             x=data_2(:,1);
             strain=data_2(:,2);

                plot(app.UIAxes,x,strain,"b");
                yline(0);
                yline(-2800); 
                hold on  
     
                % shift strain so that it positive peaks are on the y = 0 line
                [locs, pks]=peakseek(strain);
                x_peaks = x(locs);
                yshift = interp1(x_peaks,pks,x);
                strain_3 = strain - yshift;
                plot(app.UIAxes_2,x,strain_3,"b", x,yshift,"r");

           %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
            function [locs, pks]=peakseek(x,minpeakdist,minpeakh)
            % FEX : https://fr.mathworks.com/matlabcentral/fileexchange/26581-peakseek?s_tid=srchtitle
            % x is a vector input (generally a timecourse)
            % minpeakdist is the minimum desired distance between peaks (optional, defaults to 1)
            % minpeakh is the minimum height of a peak (optional)
            %
            % (c) 2010
            % Peter O'Connor
            % peter<dot>ed<dot>oconnor .AT. gmail<dot>com

            if size(x,2)==1, x=x'; end
            % Find all maxima and ties
            locs=find(x(2:end-1)>=x(1:end-2) & x(2:end-1)>=x(3:end))+1;
            if nargin<2, minpeakdist=1; end % If no minpeakdist specified, default to 1.
            if nargin>2 % If there's a minpeakheight
                locs(x(locs)<=minpeakh)=[];
            end
            if minpeakdist>1
                while 1
                    del=diff(locs)<minpeakdist;
                    if ~any(del), break; end
                    pks=x(locs);
                    [garb, mins]=min([pks(del) ; pks([false del])]); %#ok<ASGLU>
                    deln=find(del);
                    deln=[deln(mins==1) deln(mins==2)+1];
                    locs(deln)=[];
                end
            end
            if nargout>1
                pks=x(locs);
            end
            end    

[Mathieu NOE]

hello again

I have some trouble unzipping your file - does it work on your side ?

[Mathieu NOE]

maybe you could reduce the amount of data and send me only the first half , so you don't have to zip it

that should be enough data to fix the code

Answer 2

[Maria Inês]

Hi,

I had no problems but I attach a file with less data.

[Mathieu NOE]

I am tryng to find a robust method that works both for the first data file you provided and then this new one - but they show very different look

your first file (provete_15_seco.xlsx) was a clean almost triangular shape signal with no artifacts or noise. also we can see the distance between peaks are only a few samples (blue dots on the curve).

with such a clean signal peakseek would work nicely even without to give any extra parameter values about minimal peak distance and height.

the "new" data file (5x250_CF_alterado.xlsx) gives a very different figure :

• lots of data points between peaks - if I look at the negative peaks that are a bit more evident to recognize , I will see more than 1000 samples between 2 successive negative peaks, like this measurement has been made with a much higher sampling rate (or the test is done at a lower frequency)
• signal quite a bit noisy and distorded on the positive side
• peakseek will have trouble selecting the only valid peaks if we do not help him by defining an appropriate value for minpeakdist,minpeakh - now the risk is that we may have to tune those parameters depending of what signal we have , for example what distance between peaks we have and if the signal quality is like the first file or the second data file
• so it's going to be messy to tweak too many parameters depending of what the data looks like , we need something simple and robust

a zoom on the top

so my job now is to find the "magic" tool that can deal with both data sets without requiring to "tune" too many parameters....

Answer 3

[Maria Inês]

Hi,

Okay, I understand. I can't put the filter of the first data file I sent in this second file because the data is different.

How can I help? What kind of parameters do you need to know to get the magic formula?

If it's simpler, you can just find the magic formula for these data (5x250_CF_alterado.xlsx) but with the same goal, ie take the peaks and set them to zero.

There is another issue that I don't know if you've seen, Ch2 will have symmetrical values to that of CH1. The filter must also be able to make this adjustment when the peak values are negative and positive.

I don't know if I made myself clear...

Any questions, just say.

[Mathieu NOE]

I was about to ask if you have some more data files to share , so I can see what kind of signals we have , but in the mean time i figured out / tested other options and , at least for the two files we have today , I could find a simple and more robust solution

basically you only need to replace peakseek with another function that is better working on noisy signals (here comes peakfinder , see attachment)

so it seems to work for both data files , and or the last file, I tested on channel 2 and 3 and it's ok

sorry if I have a bit messed up your code , but I think you can easily copy / paste the new lines , or simply uncomment your lines I have commented

also I beleive it is better to remove NaN values (they are generated by readmatrix when you have header lines in your excel file but you don't ask readmatrix to ignore them)

ok, so finally this is the new code , please have a ook and try it and let me know !

if you want to send me some more data files , I can also continue to test the code on my side

% [name,path]=uigetfile('.xlsx','Select a file'); % your code

% my code for fast testing
path = pwd;

% files i want to test
% name = 'provete_15_secoB.xlsx'; % 1st file
name = '5x250_CF_alterado.xlsx'; % 2 nd file (test Channel 2 and 3 !)
            
            % main code
             file=fullfile(path,name);
             data_2=readmatrix(file,"Sheet","Untitled");
             x=data_2(:,1);
             strain=data_2(:,2); % choose which strain channel you want
             
             % better remove NaN's first (they come from excel file header lines)
             i1 = ~isnan(x);
             i2 = ~isnan(strain);
             i3 = i1&i2;
             x=x(i3);
             strain=strain(i3);      
             
%                 plot(app.UIAxes,x,strain,"b"); % your code
                plot(x,strain,"b");
                
%                 yline(0); % your code
%                 yline(-2800); % your code 
                hold on  
     
                % shift strain so that it positive peaks are on the y = 0 line
                [locs] = peakfinder(strain); % more noise resistant than peakseek
                yshift = interp1(x(locs),strain(locs),x);
                plot(x,yshift,"k");  % just for fun (debug) 
                strain_3 = strain - yshift;
%                 plot(app.UIAxes_2,x,strain_3,"b", x,yshift,"r"); % your code
                plot(x,strain_3,"r");
                hold off
                

[Mathieu NOE]

forgot the attachment (peakfinder )

now you have it !