Answer to Jason Rutherford

Hello Jason,

I am using mtspec the Matlab version from (German prieto web page) basically is the same, so I compute the spectrogram with an overlap of N-1 and time-bandwidth of 4 with a sample frequency of 50 and it works pretty well. Much better that welch method.

For the nfft I would use just 3 sec I mean nfft=50*3 samples.

I hope this going to help you.


Thanks for your response. Now, the mtspec output in Python, which is probably similar to your output in Matlab, returns frequencies and corresponding spectra (counts/Hz). In your workflow with mtspec, is it possible to create a spectrogram (Frequency vs time plot) from the outputs presented in the periodogram (Frequency vs. PSD)? I’ve come across a few papers that say they apply the multitaper spectral estimation method but display their spectral estimates in the form of a spectrogram, rather than a periodogram. Thoughts?