PeakFinder

Quickly finds local maxima (peaks) or minima (valleys) in a noisy signal.

Description

This function quickly finds local peaks or valleys (local extrema) in a noisy vector using a user defined magnitude threshold to determine if each peak is significantly larger (or smaller) than the data around it. The problem with the strictly derivative based peak finding algorithms is that if the signal is noisy many spurious peaks are found. However, more complex methods often take much longer for large data sets, require a large amount of user interaction, and still give highly variable results. This function attempts to use the alternating nature of the derivatives along with the user defined threshold to identify local maxima or minima in a vector quickly and robustly. The function is able to correctly identify the major peaks on a 1.5 million data point noisy sum of sinusoids in under a second as is shown in the example in the code comments.

Please don't hesitate to comment or contact me if you have suggestions about improvements that could be made to this function.

• MATLAB release: MATLAB 7.8 (R2009a)
• Other requirements: None

[Morvan]

 

[Andres]

I was happily using peakfinder until I had severe trouble with a dataset beginning with repeated values.
I've boiled it down to a short example:

t = [49 49 54 49 -57 46 -96 -10 39 0];
startI1 = 1;
[Loc1, Peak1] = peakfinder(t(startI1:end),20);
startI2 = 2;
[Loc2, Peak2] = peakfinder(t(startI2:end),20);

figure
plot(t,'b.-')
hold on
plot(Loc1-1+startI1,Peak1,'ro')
plot(Loc2-1+startI2,Peak2,'go')

Problems:
1) The first call returns a wrong result (Loc1 = 2, Peak1 = 49).
2) The second call returns all three peaks, but the last one appears twice (Loc2 = [2 5 8 8].', Peak2 = [54 46 39 39].').

I hope these issues can be fixed as peakfinder works promisingly fast.
A minor suggestion: if the input is a row vector the output should be a row vector, too.

Regards
Andres

[Andres]

The bugs were fixed quickly. Thanks Nate!

[antonio Acampora]

thanks

[Nicholas]

Does anyone know of a comparable code that finds extrema in 2D or 3D data?

[David]

fantastic code, very effective even with highly noise data.

notably, it might be worth including an histogram based thresholding for a more optimal threshold selection.

[Tim]

I think I found a mistake in the code. Or at least something strange:

if foundPeak && (x(ii) > peakMag(end) || leftMin < peakMag(end)-thresh)

The second part, x(ii) > peakMag(end), is rather strange, since peakMag(end) will ALWAYS be equal to zero, since peakMag was preallocated with zeros. I guess this is a remainder from the time before you preallocated peakMag for speed.

I replaced the mistake with the following:

if foundPeak && (x(ii) > peakMag(cInd-1) || leftMin < peakMag(cInd-1)-thresh)

The strange thing is that the results luckily stay exactly the same, which is good but strange at the same time. There is probably some redundancy in the code.

In any case, your method was more or less what I was looking for. Just need to make some small adaptations to it for my purpose, of course leaving all credits to you. Thanks!

[Nate Yoder]

Thanks for catching that Tim. You were exactly right, the redundancy is an artifact from when I preallocated the matrix for speed. However, I essentially replaced this statement with the leftMin variable which is why you got the same results. The end effect is that the second part of that conditional can simply be eliminated.

Thanks again for your help and an updated version with the addition of a user defined threshold should be available shortly.

[Elisa]

Hi. I need to find peaks in a waveform and calculate the inter-pulse interval and the pulse rate from the position of the peaks. My data are sounds recorded at 24000Hz and saved in .wav files. This code seems to be doing exactly what I want but since I’m a matlab newbie I’m not quite sure how to input my data. Do you have any suggestions?
 I hope this is the right place to post this message, otherwise I apologise.

[Estella Liu]

 

[Robert]

 

[mohammad]

 Really nice

[Diana]

Hi, I find this routine very good and useful. However, for my problem I would need the start-end information about each peak so I could select just the peaks and work with them, How can I get this information?

Thanks in advance