How to isolate peaks from a noisy power spectrum

I have a noisy power spectrum with several distinguishable peaks. Does anyone know of an automated way to isolate the peaks in an automatic way? I have to do this for multiple power spectrums.

Answer 1

[Star Strider]

The findpeaks (link) function comes quickly to mind. Also, the islocalmax (link) function, introduced in R2017b.

[Georgios Tertikas]

thank you for your answer. however these functions do they work in case of a noisy signal? I want to isolate the higher peaks in a graph where there are some much smaller peaks.

[Star Strider]

The findpeaks function has more name-value pair options to deal with noise than islocalmax, in my experience. Without your signal (and the code you use to calculate the spectrum), it is not possible to be more specific.

[Georgios Tertikas]

In order to caluclate the spectrum i used fast fourier transform data in an already given data set. Then It is raised ^2 to give the power spectrum. I hope this covers some things.

[Star Strider]

‘I hope this covers some things.’

It does not add any useful information.

Note that you need to square the magnitude (using the abs function) of the normalised fft output to calculate power.