Really enjoyed the lecture but all the images and videos were quite distracting. I found they actually ended up reducing the impact of what was being said.
I saw this for the first time a few years ago, and I thought it was brilliantly done. I was really glad to see it was still findable. I admire your ability to explain such a relatively complex topic in such an approachable way. If all teachers were so humble there would be a lot more scientists and engineers in the world.
Great Presentation. It would be great if you add more videos on you tube on this topic. I am interested to know how wavelets can adaptively change their windows shape to handle changes in non-stationary signal.
I mean how a wavelet would know that in next millisecond the signal under test is going to have higher or lower frequency so that it change itself accordingly to balance t-f resolution issue.
Really nice explanation. I use Kyma (Symbolic Sound Corp.) which uses the STFT to analyze wave forms to perform spectral morphing between 2 signals. But in order to morph from 1 sound to another, we need 2 signals that are positioned in the same time domain as 1 another (time dependent). This uses a large amount of DSP in the Capybara. I wonder if the Wavelet Transform might allow this time dependent aspect more flexibilty so that live morphs independent of time resolution within a performance?
This is something you probly checkout through published articles. Someone probly has gone through this type of analysis examining the best topology. There are other wavelets besides Daubechies. They should be examing orthangonality, symmetry, and other topology components of the waveform.
Are you just TRYING to write something intelligent? One of the main concepts of wavelets is the Topology of the mother wavelet. Thats why it so special and why it works better than a using sinoids. For instance, using morlet wavelets work great for neuronal signals because of how well it works with the topology of the an action potential.
STFT are good and sometimes thats all you need. You dont need anything fancy. HOWEVER, Using a STFT you have to know what the size of your window is going to be! Wavelets you dont! It is obvious you didnt get the symbolism in the beginning and end of the video. One of the sayings with the wavelet transform is that you can see the forest and the trees. This saying implies the use of the multiple windows that abide to the Heisenberg Uncertainty Principle .
hang on you plum, you can just have a bunch of different window sizes and call them scales just like in wavelets - you just use nice simple waves rather than some randomly plucked wavlet function. I refer this debate to Occam
But how can modify the parameters of your window on fly that fits the t-f resolution requirement, and can be set to meet the non-stationarity of the signal under observation.
As far as STFT is concerned, it is the simplest tool to analyze the no-stationary signal; but does not always yield the desired results. I have developed a technique that performs better than STFT but it not adaptive like wavelets. I wanna look into wavelets in more depth.
Additionally, if you really want to get technical, there are nonlinear wavelets but thats a whole new can of worms. STFT is quite inferior to wavelets but sometimes you dont need something so technical to solve the problem at hand. Its obvious that you really cant handle something so cumbersome.
P.S. Even though your opinion is unreliable and invalid..I value it. THANK YOU but try cracking a book!
@dibbuck: Well, STFT has clear drawbacks like the fourier uncertainty principle. That's why wavelet transformation was developed.
And why is the possibility of the arbitrary choice of the mother wavelet bullshit? That's one of the advantages of wavelets, I would say. This and the possibility to overcome the uncertainty problem.
Maybe you mean FTFT that's useful to rotate signals in the time-frequency domain and has its own use.
There are sound processing applications that use wavelets..However, there isn't much out there because FFT's processing speed is much much higher than using wavelets. One of the down sides of the use of wavelets particularly when processing high Sampling Frequencys and the duration of bouts.
cooool
pmsutube 1 week ago
MUST WATCH....for electrical engineer....
rkchinmay 1 month ago
very nice
AlwaysAndForeverPFC 1 month ago
This has been flagged as spam show
wavelet transform is the future to time-frequency decomposition try it you won't regret it =)
NaRoChIn 6 months ago
Really enjoyed the lecture but all the images and videos were quite distracting. I found they actually ended up reducing the impact of what was being said.
meme2342 9 months ago
I saw this for the first time a few years ago, and I thought it was brilliantly done. I was really glad to see it was still findable. I admire your ability to explain such a relatively complex topic in such an approachable way. If all teachers were so humble there would be a lot more scientists and engineers in the world.
justinhabit 1 year ago
Thanks, a great presentation!
rouzbehmaani 1 year ago
My darling boyfriend,
I'm making my own video like this. ;)
xo
TheJeneferTaylor 1 year ago
Well done!
srki013 1 year ago
Ingrid Deebahshees. :)
Daubechies is French. Try "Daw-beh-shee" as a better rendition of the pronunciation, with the last syllable stressed :)
And Fourier is pronounced "Foo-ree-yeh", again, last syllable stressed.
Great introduction to wavelets, by the way. Thumbs up.
albedoshader 1 year ago
@albedoshader Daubechies is a Belgian and American (naturalised 1996) citizen. She speaks French but is in no way French.
naxelas 1 year ago
@naxelas: I didn't mean she's Freanch but her surname is. :) Sorry for my imprecision. I should have said: the name "Daubechies" is French.
albedoshader 1 year ago
@albedoshader All right then ;-) It indeed sounds really French.
naxelas 1 year ago
Best tutorials i have ever seen.......
ashankbeenu 1 year ago
This is the most intuitive tutorial in wavelet analysis. Excellent.
paucarre 2 years ago
Great Presentation. It would be great if you add more videos on you tube on this topic. I am interested to know how wavelets can adaptively change their windows shape to handle changes in non-stationary signal.
I mean how a wavelet would know that in next millisecond the signal under test is going to have higher or lower frequency so that it change itself accordingly to balance t-f resolution issue.
AnnManMS 2 years ago
Really nice explanation. I use Kyma (Symbolic Sound Corp.) which uses the STFT to analyze wave forms to perform spectral morphing between 2 signals. But in order to morph from 1 sound to another, we need 2 signals that are positioned in the same time domain as 1 another (time dependent). This uses a large amount of DSP in the Capybara. I wonder if the Wavelet Transform might allow this time dependent aspect more flexibilty so that live morphs independent of time resolution within a performance?
lysergicwindow 3 years ago
Nick1Nap,
Is there specific Daubechies to heart rate variability analysis using wavelet tranform?
lperandini 3 years ago
This is something you probly checkout through published articles. Someone probly has gone through this type of analysis examining the best topology. There are other wavelets besides Daubechies. They should be examing orthangonality, symmetry, and other topology components of the waveform.
Nick1Nap 3 years ago
mother wavelet choice is subjective and hence bullshit. Get off the fashion bandwagon and embrace stft
dibbuck 3 years ago
Are you just TRYING to write something intelligent? One of the main concepts of wavelets is the Topology of the mother wavelet. Thats why it so special and why it works better than a using sinoids. For instance, using morlet wavelets work great for neuronal signals because of how well it works with the topology of the an action potential.
Nick1Nap 3 years ago
STFT are good and sometimes thats all you need. You dont need anything fancy. HOWEVER, Using a STFT you have to know what the size of your window is going to be! Wavelets you dont! It is obvious you didnt get the symbolism in the beginning and end of the video. One of the sayings with the wavelet transform is that you can see the forest and the trees. This saying implies the use of the multiple windows that abide to the Heisenberg Uncertainty Principle .
Nick1Nap 3 years ago
hang on you plum, you can just have a bunch of different window sizes and call them scales just like in wavelets - you just use nice simple waves rather than some randomly plucked wavlet function. I refer this debate to Occam
dibbuck 3 years ago
But how can modify the parameters of your window on fly that fits the t-f resolution requirement, and can be set to meet the non-stationarity of the signal under observation.
As far as STFT is concerned, it is the simplest tool to analyze the no-stationary signal; but does not always yield the desired results. I have developed a technique that performs better than STFT but it not adaptive like wavelets. I wanna look into wavelets in more depth.
AnnManMS 2 years ago
Additionally, if you really want to get technical, there are nonlinear wavelets but thats a whole new can of worms. STFT is quite inferior to wavelets but sometimes you dont need something so technical to solve the problem at hand. Its obvious that you really cant handle something so cumbersome.
P.S. Even though your opinion is unreliable and invalid..I value it. THANK YOU but try cracking a book!
Nick1Nap 3 years ago
P.P.S. Profanity is a weak attempt of a feeble mind trying to express itself
P.P.P.S. Obviously, this applies to you!
Nick1Nap 3 years ago 2
time on your hands perchance?
dibbuck 3 years ago
@dibbuck: Well, STFT has clear drawbacks like the fourier uncertainty principle. That's why wavelet transformation was developed.
And why is the possibility of the arbitrary choice of the mother wavelet bullshit? That's one of the advantages of wavelets, I would say. This and the possibility to overcome the uncertainty problem.
Maybe you mean FTFT that's useful to rotate signals in the time-frequency domain and has its own use.
albedoshader 1 year ago
nice video, is there a reason why don't any sound processing applications use a wavelet transform instead of fft?
shep59 3 years ago
There are sound processing applications that use wavelets..However, there isn't much out there because FFT's processing speed is much much higher than using wavelets. One of the down sides of the use of wavelets particularly when processing high Sampling Frequencys and the duration of bouts.
Nick1Nap 3 years ago
really good
rand69m 3 years ago
Very very good!!
diegokillemall 3 years ago
Man, it really helped me. Thank you so much
preeti06rao 3 years ago
Thanks so much for this video. it really helped me understand wavelets!
poojashail 3 years ago
You are welcome...glad to hear it helped! If you have anything i may help you with. I would be glad to help.
Nick1Nap 3 years ago