The signal I'm studying has fundamental frequencies of 20 and 60 cycles per minute (shown in the Periodogram graph). It is straight forward to extract the peaks in the time domain belonging to the 20 cycles/minute frequency (both are circled), however the peaks associated with the 60 cycles per minute frequency can be noisy.

Time Domain

6 seconds of the time domain signal

Frequency Domain (per minute not per sec)

Both signals in the time domain will ALWAYS overlap so low pass filtering isn't completely necessary but I'm just curious to know if there's a better way of extracting the 60 cycles per minute peaks (aside from peak-to-trough analysis) using frequency domain information?


  • $\begingroup$ Are you after the locations of the peaks as it was the only signal or peaks in the composed signal? $\endgroup$
    – Royi
    Commented Jun 26, 2019 at 6:48

1 Answer 1


You can figure where the peaks would be if the other tone wasn't there by an accurate estimate of the phase. This location will be a rougher estimate in the presence of the other tone. When you add a slanted signal (say a line) to a mode (a peak), it will shift the location of the peak somewhat. Since you are interested in the higher frequency tone, the lower frequency one is going to look like a broad hill in comparison which is going to be less interfering than if you were doing it the other way around. (I'm using the word "tone" broadly, meaning a pure sinusoidal signal.)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.