How to detect peaks above a given threshold value?

findpeaks() can detect multiple peaks if there is/are change in direction(s) even if the values are above the specified threshold (see arrows in the attached pdf file).

J=a vector;
findpeaks(J,'MinPeakHeight',Threshold);hold on;
yline(Threshold,'--','LineWidth',3);
hold off

[madhan ravi]

Is your threshold 20 ?

[Sudip Paudel]

Hi Madhan, thank you for asking.

In this specific example, I was using 5, but this threshold can vary.

[madhan ravi]

What happens when you change it to 20?

[Sudip Paudel]

Hi Madhan, we will not detect other peaks if I change the threshold value to 20. My goal is to count it only once if the signal goes above the threshold value until it goes below the specified threshold.

Answer 1

[Rashedur Rahman]

Hi Sudip.

The function 'findpeaks' can take several input arguments for finding peaks according to criteria. You can use multiple arguments to adjust your peak finding algorithm. I will suggest you to read the description of 'findpeaks'. ('https://www.mathworks.com/help/signal/ref/findpeaks.html').

Another way is to apply a filter to smoothen the signal and then apply 'findpeaks'.

Answer 2

[Image Analyst]

Try flattening the signal by subtracting that red trend line. Then set the values of the signal below the threshold to zero and then call findpeaks()

smoothedSignal = movmean(signal, 101);
flattenedSignal = signal - smoothedSignal;
flattenedSignal(flattenedSignal < someThreshold) = 0;
[peakValues, indexesOfPeaks] = findpeaks(flattenedSignal);

Attach your data in a .mat file if you need more help.

[Sudip Paudel]

Thank you everyone for your answers.