How can I find a minimum only before a peak has been detected?

I have the following graph in which I used function findpeaks and MinPeakHeight to find the peaks that I'm interested in.

[pks,locs]=findpeaks(filter_sgo,'MinPeakHeight',limiar,'MinPeakDistance',minpkdist);

Now I want to find the local minimums but only before the peaks that I have already detected (red points of the picture). I know I have to invert the signal and use findpeaks again, but I can't figure out how to only detect before a peak.

Thanks a lot.

[Jan]

Do you mean the the interval between 0 and the first peak "1" or before each peak?

[jonas]

Could you perhaps draw some of the points you would like to detect? Include some of the more ambiguous ones, like the local minimum prior to '3' and '22'

[Ana Silva]

I updated the picture. The graph is a recording of emg, so i'm only interested in detecting minimums before a peak that corresponds to a contraction. Minimums like the ones before 9, 21 and 22 I don't want to detect because they don't correspond to the beggining of a contraction. But I suppose this is not a problem because I can do the same I did for the peaks, and only detect a minimum under a certain value.

[jonas]

Well from the figure it seems all your points are located around y=0-5 and you are not interested in higher ones. Wouldn't the simplest solution be to find the first value below a threshold, stepping backwards from each peak?

[Ana Silva]

I have the location of the peaks and of the minimums. I tried to do as you said:

%Location of peaks  
[pks,locs]=findpeaks(filter_sgo_short,'MinPeakHeight',limiar,'MinPeakDistance',minpkdist); 
%Location of minimums
[peaks_min,loc_min]=findpeaks(DataInv(1,:),'MinPeakHeight',50,'MinPeakDistance',2000);

for j=1:length(locs)
%go from the peak to 2000 points backwards
for i=locs(1,j):locs(1,j)-2000
if loc_min(1,:)< locs(1,j) && loc_min(1,:)>locs(1,j)-2000
    loc_min2=[1,loc_min];
end
end
end

loc_min2 is not being created at the end of the loop. Do you know what I'm doing wrong?

[jonas]

If you post the data (or part of the data) and your code, then it would be easier to help you. It would also be easier to post a good answer. Anyway, I'm guessing this line is incorrect

if loc_min(1,:)< locs(1,j) && loc_min(1,:)>locs(1,j)-2000

you are never using i in your loop, perhaps : should be replaced by i?

[Folakemi Omotoye]

Hi Anna, How did you get this kinda plot. I want to get a similar plot to this

Answer 1

[jonas]

What you are looking for seems to be the beginning of the peak, which is not necessarily a valley. A simple method to approximate those values would be to count backwards from each peak, and find the location when the signal intercepts a predefined threshold. Here's a simple example. You may recognize the peaks from the documentation on findpeaks.

x = linspace(0,1,1000);
Pos = [1 2 3 5 7 8]/10;
Hgt = [3 4 4 2 2 3];
Wdt = [2 6 3 3 4 6]/100;
for n = 1:length(Pos)
    Gauss(n,:) = Hgt(n)*exp(-((x - Pos(n))/Wdt(n)).^2);
end
PeakSig = sum(Gauss);
%%example starts here
[pks,locs] = findpeaks(PeakSig);
locs=[1 locs];
thres=0.5; %set threshold
win=diff(locs)
mins=nan(size(win))
for i=1:numel(win);
    ind=find(PeakSig(locs(i):locs(i+1))<thres);
    if ~isempty(ind)
    mins(i)=max(ind)+locs(i);
    end
end
mins(isnan(mins))=[];
figure;hold on
findpeaks(PeakSig)
plot(mins,PeakSig(mins),'rx')

[Ana Silva]

thank you very much! i've managed to do what I wanted

[jonas]

Great!

[Image Analyst]

Or (what I think might be better) is to descend from the peak until you're less than the threshold, and keep going until the values start to turn around and increase again.