How can I set a descend order finding peaks to my graph ?

I have a graph that takes the values from first maximum peak point and plots it.Now,how shall i introduce a threshold or descend order from that point,so that it find the next highest peak ?

load('signal')
load('t')
S(:,1)=[ 5 15 35 45 5 15 35 45];
S(:,2)=[ 5 15 15 5 45 35 35 45];
 for k=1:1:length(S) 
[a(k),b2(k)]=max(abs(hilbert(signal(k,:)))); % first peak amp & index
    
 [PkAmp, PkTime] = findpeaks(abs(hilbert(signal(k,:))),t); 
  
  peaks(1,1)=a(k);
  peaks(2,1)=t(b2(k));
  
 j=2;
 for i=1:length(PkTime)   
 if PkTime(i)>t(b2(k))
     peaks(1,j)=PkAmp(i);
     peaks(2,j)=PkTime(i);
     j=j+1;
 end
 end
figure;  
    plot(t,(signal(k,:)),t,abs(hilbert(signal(k,:))),peaks(2,1),peaks(1,1),'ro')
    hold on;grid on
    plot(PkTime, PkAmp, '^r', 'MarkerFaceColor','r')
    xlim([0 0.0006]);ylim([-inf inf])
    hold on,plot(peaks(2,3),peaks(1,3),'ko')
    
 end
 
 

[darova]

I don't understand

[Ramesh Bala]

The code calculates the peaks only from the Hilbert maximum.No peaks before Hilbert max point is considered in the present code. The query is how to calculate the second highest peak value . As shown in the figure that should be the second peak and not the peak which is next to Hilbert max point

[Ramesh Bala]

In other words I would like to take only the points that are at sharp peaks, not any peak values which are in a slope or down slope.

[darova]

What if just use findpeaks() two times?

[y, ix] = findpeaks(origin_data);
[y1,ix1] = findpeaks(y);
i = ix(ix1);
plot(time,origin_data)
hold on
plot(time(i),origin_data(i),'^r')
hold off

[Ramesh Bala]

Thanks Darova for the comment.

It doesn't work as it didn't made any envelope over it.

Answer 1

[Akira Agata]

How about combining envelope and findpeaks functions?

The following is an example.

% Load data
load('signal.mat');
load('t.mat');

% Calculate envelope and detect it's peaks
ul = envelope(signal(1,:));
[pks,locs] = findpeaks(ul,t,'MinPeakProminence',0.5e-3);

% Show the result
figure
plot(t,signal(1,:))
hold on
plot(t,ul)
scatter(locs,pks,'rv')
legend({'Original data','Envelope','Peaks'},'FontSize',12)

[Ramesh Bala]

Thanks Akira Agata for the answer.

Indeed it looks promising.But if you check the figure it didn't took the next sharp peak as 3rd put took in another period peak.

Also,this is one signal,but when I do it in a loop for all the 8 signals,it even takes the value before the highest envelope point.I tried to increase ,decrease the minpeak prominence value to 0.4e-3-0.7e-3 range still not satisfactory.I have to atleast check if this works for 6 peaks.The perfect one was figure6,figure 5,8 takes the peak values before the envelope max. Attaching figure1 and figure6,7,8 from the code.

Reattaching the code with the loop

 % Load data
load ( 'signal.mat' );
load ( 't.mat' );
% Calculate envelope and detect it's peaks

for k=1:1:length(signal)
ul = envelope (signal (k, :));
[pks, locs] = findpeaks (ul, t, 'MinPeakProminence' , 0.5e-3);
% Show the result
figure
plot (t, signal (k, :))
hold on
plot (t, ul)
scatter (locs, pks, 'rv' )
legend ({ 'Original data' , 'Envelope' , 'Peaks' }, 'FontSize' , 12)

end

With minpeakprominence value =0,5e-3

Peak 2 is correct ,but 3.4 seems not identifed and 5thor 4th valyeis identifed as 3rd.

Correctly identified peaks-fig 6 - all 6 peaks clearly shown

Fig 5: negative peak identified

Fig 8:negative peak identified

[Ramesh Bala]

I tried to add these to the code

[PkAmp, PkTime] = findpeaks(abs(hilbert(signal(k,:))),t,'MinPeakDistance',7e-5,'MinPeakHeight',0.2*a(k));

But,still it couldn't find the first 6 peaks.

[Akira Agata]

Then, how about the followings?

To detect peaks more precisely, I believe you need to tune parameters in findpeaks and envelope function and/or add more options to the findpeaks function.

% Load data
load('signal.mat');
load('t.mat');

% Calculate original peaks
[pks0,locs0] = findpeaks(signal(1,:),t);

% Calculate envelope and detect it's peaks (max: 6 peaks)
ul = envelope(signal(1,:),10,'peaks');
[pks,locs] = findpeaks(ul,t,...
  'MinPeakProminence',0.1e-3,...
  'NPeaks',           6);

% Find nearest original peak for each envelope peak
D = abs(locs0 - locs');
[~,pt] = min(D,[],2);

% Show the result
figure
plot(t,signal(1,:))
hold on
plot(t,ul,':')
scatter(locs0(pt),pks0(pt),'rv')
legend({'Original data','Envelope','Detected Peaks'},'FontSize',12)

[Ramesh Bala]

Thank you ! it works well for most signals !

I would like to know how to set that Minpeak prominence ,could you elaborate how to identify it and its significance.

[Ramesh Bala]

I would like to know how can I further choose different peak values and store them in a separate variable.

Currently pks0(pt) and locs0(pt) is stored as a variable in a linear way according to signal numbers.

But,is there a way using brush option or something to select single peak from a figure and store it as (1,1) and then another peak from another plot and store as (1,2)