Thank you so much, appreciate the feedback.

Thanks for your feedback. This topic has been discussed here dsp.stackexchange.com/questions/78514/wavelet-scattering-properties-implementation/78515#78515. An example is also available here where the authors train a convolutional network to invert the scattering transform see www.kymat.io/gallery_2d/regularized_inverse_scattering_MNIST_torch.html Read this paper if you trying to implement GANs: arxiv.org/pdf/1805.06621

@@AlaphBeth Thank you very much.

Check this book link.springer.com/book/10.1007/978-3-642-56702-5

Check this book link.springer.com/book/10.1007/978-3-642-56702-5

Hi there, thanks for your feedback. Have a look at this one link.springer.com/book/10.1007/978-3-642-56702-5 The video is the result of books, papers and work experience, so you may or may not find everything in the books. However, this specific book is one of the best IMHO.

@@AlaphBeth Thank you very much! This helps a LOT!

Thank you for your feedback. I would recommend this book link.springer.com/book/10.1007/978-3-642-56702-5. However, there are many more out there. For me, it was mostly the papers and the book chapter I read from here and there plus work experience.

Thank you very much!

Thanks for your feedback. That has been actually done by many researchers, see for example www.hindawi.com/journals/mpe/2019/1340174/ Or this one www.researchgate.net/publication/334519126_LSTM_with_Wavelet_Transform_Based_Data_Preprocessing_for_Stock_Price_Prediction

Thank you so much for your feedback.

Hello, thanks for your inquiry. The database associated with this demo is available here www.rami-khushaba.com/biosignals-repository, it’s the very last one at the bottom of the page. This demo was prepared around 2011 to support our publication. Calibration is always an issue with EMG pattern recognition. Look at this paper for some ideas ieeexplore.ieee.org/document/6737313

Thank you just one more question was the model also trained on the emg of the person thats in the video performing the gestures?

@jomfawad9255 yes, model was trained on the subject data and then put in test in real time on his real-time data. Be aware though, if the person gets fatigued or if the sensors locations change or shift during real-time then even if you include his/her data in training then the performance would still be impacted heavily by electrodes shifts.

@@AlaphBeth For such software do you recommend to train software on the person's emg signals plus a large database of people performing gestures with their emg signals or just the person himself in enough? thank you

Thanks for the feedback, much appreciated.

hello hassansaad, "I am currently working on a thesis similar to your task. May I see your guidelines on how to understand what DWT, signal decomposition, and db4 are?"

Thank you for your feedback. This is the way it was designed originally as the goal here is to separate the high frequency fluctuations from the rest of the signal. We keep decomposing the low pass till we reach the frequency range of interest or to the point we can’t decompose further beyond (criteria to check that is known). However, wavelet packet goes further and says why only decomposing low pass side and let’s do that for the high pass side too and by that providing a better picture of the time frequency contents of the signal.

Thank you for your feedback. The answer is not the sum of the individual values, as you just need to think multidimensional rather than single dimension. The answer for that lower left bin is a count of the number of samples from feature 1 that falls within the range of the first bin while the samples in the second feature fall within the range of the first bin of that feature. In that example it is given as how many samples from feature 1 fall within the range of 0.1 to 0.34 while at the same time the samples from feature2 had values within the range of 1 to 1.59. Think of this as driving on a highway with 4 lanes. How many times did the car on lane 1 drive on a speed of 60 to 70 km for example while, at the same time, the car on lane 2 had a speed between 80 to 90km for example. When you do individual cars you just look at car on lane 1 individually from other cars and count number of times it drove on a specific speed regardless of other cars on other lanes.

Hi Hassan Let’s take this one by one. EEG is a non-stationary signal. By using FFT on EEG, you may lose some info and you have to work on small segments of the EEG signal during which one can assume that the EEG is stationary. FFT can though show you the power spectrum with frequency on the x axis and power on the y axis, that is the contribution of every single frequency. On the other hand, DWT can deal with non stationary signals like EEG. It usually chops the frequency spectrum into bands or segments and keep dividing these bands into smaller portions that allow us to zoom at specific events of interest. In the video, I already showed you how to extract the EEG bands related features. So if you ask can you plot EEG bands then I would say watch the video again as you obviously missed this part. Your second question, DWT is not a frequency analysis tool like FFT, it’s a time frequency analysis tool or time scale. Hence you can get localised info in time and frequency. For your question in which one to use, I would say use whatever works the best for your example.

Thanks for your feedback. With every decomposition step you have a filtering process and a down sampling step. The down sampling step will result in less samples being kept as you keep decomposing. This downsampling applies to left and right sides of the generated nodes. As a result, the number of samples left after each decomposition step reduces on both sides. About your question on wavelength, not sure why you assume the left most to have the greatest duration if you keep downsampling! Check the book titled Ripples in mathematics, it explains it all in an excellent way link.springer.com/book/10.1007/978-3-642-56702-5

Thank you for your feedback. The choice of the method depends pretty much on the platform in which you are applying the baseline wandering methods. If you are running these kind of analysis on a PC then you can use whatever method there and specifically explore the deep learning approaches as the literature is promoting these methods. However, if you are considering running these analyses on a wearable device, then it’s all about the computational power of your wearable device that will determine what method(s) you can use. As for the selection of the methods themselves, I have explored a number of these in the video specifically for ECG signals, and there could be more methods out there, but it all depends again on what you want specifically, that is to extract some features without caring much about the shape or pattern or you want very nice:clean representations for exploring:plotting or other purposes.

Thank you, appreciate the feedback.

Thank you so much for your feedback, much appreciated.

Thank you for your feedback, much appreciated. I tried to draw an intuitive example that a new starter in the field can relate to and couldn’t find something easier from the real world rather than the train example :)

​@@AlaphBeth Your video is impressively detailed and the effort you've put in is clear. Thank you! The train metaphor was a decent attempt but it was a bit confusing. Have you thought about using audio/music as an example instead? Consider a song, where each instrument contributes distinct frequency components unique to its timbre, which can be represented as Fourier constituents in a signal. Instruments don't always play throughout the entire song - they enter and exit. Similarly, our signal's frequency content varies temporally. This variation can be visualised in a spectrogram, correlating to temporal analysis.

@@dreamdrifter Good one 👍 For me the idea of train/people seemed more appealing at the time as you get to see people sitting in the different seats but you can tell when they went inside the train (if you are sitting already), and from outside the train you don’t know what seats they will take without being inside, which paved the way for time frequency view (standing near the door). I guess with some nice presentation the idea of song would look great in this topic.

@@AlaphBeth Indeed. However, it is not clear how the distribution of the people standing on the platform has any temporal element - I perceive that as a spatial distribution. And to be difficult, if I am sitting on the train I can turn my head towards the door and see who is coming in and out. I understand the analogy assumes otherwise but this fact is likely to cause some cognitive dissonance (at least, it did for me)

Thank you for your feedback, much appreciated. When you have some time, it will be great if you can share with me what went wrong and the misconceptions through an anonymous email through my website www.rami-khushaba.com/contacts.

Thanks for your feedback. Yes, the repo link is shown on slide 13, github.com/mitbal/py-bwr

The number of scattering coefficients depends on the values of the SignalLength, InvarianceScale, and OversamplingFactor properties of the scattering framework SF. Specifically, len = 1+fix((sl-1)./2^(cr-osfac)); sl = SignalLength cr = criticalResolution osfac = OversamplingFactor

Hello, this video is from 11 years ago, so as much I would like to help you here, I honestly don’t have any code here for grove EMG data collection.

Hello, thanks for your feedback. About your question: for the case of two variables, the exterior envelope of the two circles is H(X,Y) that is the entropy of the two variables. Mutual information, I(X;Y) is given by the overlapping region between the two circles. So you start from entropies, and the joint areas are the portions that one variable knows about another and hence reducing entropy - this is mutual information. The same goes for the case of three variables. I(X;Y;Z) is the joint part between the three circles and the outside envelop for the three circles is the entropy.

Thank you, ah yes I mixed up the entropy and the mutual information. It’s clear now, thanks!

Thank you for your feedback. If you mean the detrend function in Matlab, that one is based on curve fitting, I.e., fitting a curve to the data and then subtracting that curve away. This is not explicitly looking at filtering a range of frequencies like the FFT, classical filters, or wavelet approaches. So the methods in the video look at frequency domain analysis or time-frequency analysis versus detrend that does curve fitting. On the other hand, the deep learning approach learn how to act as a robust filter based on how much data you show it (can be thought of on both sides).

@@AlaphBeth Hi, dr.. If using DWT db4 level 10 to ECG Signal to removing noise Fs= 360 baseline wander (0 - 0.5hz) = [a10+d10+d9+d8+d7+d6+d5+d4+d3+d2+d1] =for feq between (0- 0.35hz) and for power line interference (50 / 60 hz) PL= [a4+ d4+ d3+d2+ d1], is the correct dr.

@@monashaaban2337 there is no correct and wrong here, it’s about what works the best for your signals. Just run your code and see how good it is denoised. For the baseline wander that range seems to be within the limits so it should give you some good approximation there. For the power line interference, 4 levels of decomposition will give you ranges of nearly 22.5Hz and that is a bit wide range if you want to filter 50Hz. As you are trying to do multiple ranges of frequencies, how about you try the same idea in the video but with wavelet packet transform? That will give you more control over the ranges.

@@AlaphBeth thank you dr.

@@monashaaban2337 happy to help anytime, keep the questions coming if you have more :)

Thank you, appreciate the feedback

Thank you for your feedback, I am glad it helped.

Here is one paper that discusses the implementation details for denoising with wavelets in real-time www.researchgate.net/publication/4174293_A_real-time_system_for_denoising_of_signals_in_continuous_streams_through_the_wavelet_transform. A more recent experiment with a DSP kit is also available here akulmalhotra.github.io/posts/2020/05/waveletecg/. Myself, I have used the DWT for EMG feature extraction (not ECG denoising) in real-time experiments, working on windows of 256ms worth of data. As for your question, a short buffer is suggested.

Thanks! I'll give this a read.