We have recently implemented a Python version of the multitaper spectrogram which you can find on Github here - github.com/preraulab/multitaper_toolbox/tree/master/python.
Assumptions: 1. EEG output(s) are representative (but are actually a reduced dataset). 2. Fourier transformation is the conversion tool to use on these reduced signals. 3. noise reduction is acceptable due to point 2. Or am I missing something?
Thanks for this video, it is super helpful and is presented very clearly that is easy to understand. It has been a major help to my thesis project, thanks again!
Thank you so much for this video! I´m taking a class about EEG/MEG this semester and this is really the first time I understood how spectral analysis works
We have recently implemented a Python version of the multitaper spectrogram which you can find on Github here - github.com/preraulab/multitaper_toolbox/tree/master/python.
Thanks a lot, Professor, for this informative video. Very well explained!! I have a question regarding the parameters of the tapers- When we chose the value of N and Δf to be 30s and 1 Hz resp., the TBW comes out to be 15 and L=29. However, the sfreqs generated from the output of the function multitaper_spectrogram(...) doesn't have the resolution of 1 Hz, instead, it has 0.02 Hz resolution. I am wondering why is it so.. Am I misunderstanding something here? Can you please clear this doubt?
There is a difference between the spectral resolution, which is the main lobe bandwidth, and the sampling interval of the frequency vector. The spectral resolution is a property of the spectral estimator, which in practice relates to the minimum observable frequency spacing between two adjacent spectral peaks. However, in computing the spectrum, we can sample values at any frequency. So in your case, the 0.02Hz is just the distance between points at which we are sampling from a spectrum with a 1Hz spectral resolution.
question: when I do a FFT of a time series with the length L and the samplerate Fs, the frequency resolution is something like Fs/L. Why not use this as the desired frequency resolution delta f? I want to utilize the multitaper method for tree ring data. I set/interpret the window size N with the help of the climate definition as weather over 25 years. With a 170 year time series I end up with a TW < 1 and thus a k of tapers of 0 or -1. Can you help?
Please give Matlab code for MTM spectrum
Thank you very much for your video, You explained very clear.
What are good python packages for multi-taper spectral analysis?
We have recently implemented a Python version of the multitaper spectrogram which you can find on Github here - github.com/preraulab/multitaper_toolbox/tree/master/python.
Great videos. Thanks
Assumptions: 1. EEG output(s) are representative (but are actually a reduced dataset). 2. Fourier transformation is the conversion tool to use on these reduced signals. 3. noise reduction is acceptable due to point 2. Or am I missing something?
Very well explain. Thank you
Thank you for the video. Hope you deliver more videos in the future.
Great this lecture is awesome. I was able to completely understand what i was curious about
Thanks for this video, it is super helpful and is presented very clearly that is easy to understand. It has been a major help to my thesis project, thanks again!
Thank you so much for this video! I´m taking a class about EEG/MEG this semester and this is really the first time I understood how spectral analysis works
Great content, Thank you so much!!!
Thanks so much for this video! I'm a maths person who's not particularly familiar with DSP or EEGs, and this helped me with both at once!
I am also seeking a good multitaper package for Python if you have that information available.
We have recently implemented a Python version of the multitaper spectrogram which you can find on Github here - github.com/preraulab/multitaper_toolbox/tree/master/python.
Thanks a lot, Professor, for this informative video. Very well explained!!
I have a question regarding the parameters of the tapers-
When we chose the value of N and Δf to be 30s and 1 Hz resp., the TBW comes out to be 15 and L=29. However, the sfreqs generated from the output of the function multitaper_spectrogram(...) doesn't have the resolution of 1 Hz, instead, it has 0.02 Hz resolution. I am wondering why is it so.. Am I misunderstanding something here?
Can you please clear this doubt?
There is a difference between the spectral resolution, which is the main lobe bandwidth, and the sampling interval of the frequency vector. The spectral resolution is a property of the spectral estimator, which in practice relates to the minimum observable frequency spacing between two adjacent spectral peaks. However, in computing the spectrum, we can sample values at any frequency. So in your case, the 0.02Hz is just the distance between points at which we are sampling from a spectrum with a 1Hz spectral resolution.
@@preraulab Thanks a lot!! I got it.
question: when I do a FFT of a time series with the length L and the samplerate Fs, the frequency resolution is something like Fs/L. Why not use this as the desired frequency resolution delta f? I want to utilize the multitaper method for tree ring data. I set/interpret the window size N with the help of the climate definition as weather over 25 years. With a 170 year time series I end up with a TW < 1 and thus a k of tapers of 0 or -1. Can you help?