3 Challenges in Signal Processing (ft. Paolo Prandoni)

This video presents 3 challenges faced by signal processing researchers. It features Paolo Prandoni, senior researcher of the IC School at EPFL.
people.epfl.ch/paolo.prandoni

[@RAJIBLOCHANDAS]

Excellent. I wish signal processing community will explore the physics of deep learning.

[@RAJIBLOCHANDAS]

I will personally try to explore that. My lecture on Gradient Descent algorithm. link: kzbin.info/www/bejne/ipOooatun5uJndE

[@sophiacristina]

Signal processing is the future!

[@loveforallbxlmannif]

“Digital signal processing is a veritable ocean. Take as many soundings in it as you will, you will never know its depth.” [Paraphrased from “Le Pére Goriot” by Honoré de Balzac, 1835] (i copied this one from dsprelated forum, a message of Rick Lyons

[@MusicEngineeer]

I'm a bit confused. When I place - say - 10 microphones into a room and let them record an audio signal, I don't get a 10-dimensional signal. I get 10 separate 1-dimensional signals, which is a very different thing - as different as an image is from a stereo audio signal - the image is truly 2D whereas stereo audio is 2-channel 1D. Number of channels and signal dimensionality seem to be conflated here, unless I'm missing something. Likewise, 10 cameras in a room would capture 10 separate 2D signals, i.e. 10 channels of a 2D signal (which could themselves actually be 3-channel data for RGB). ...or maybe 3D, if they capture video and time is taken as 3rd coordinate

[@pyb.5672]

Imagine that that you perform a partial differential equation between the spectral analysis of 2 of your microphone, and feed the resulting signal as a function of change in orientation of another of your microphone, let's say the x-axis. The signal picked up by that last microphone is on a whole other dimension. It's not just "another channel" anymore. Now imagine that you you have a compressor whose input is now the output of that last microphone, and that compressor treats all the microphones you got in the room. You can get some really, really interesting things dimensionally.

[@Bm23CC]

Deep convolutional networks is mind blowing stuff.

[@Fruitysfaction]

who is the guy interrupting all the time?

[@randythamrin5976]

same question . lol

[@arthurs3058]

Lol👍🤣🤣

[@kurchak]

😂 so random huh lol

[@daesekang8147]

lol yeah i feel like it was unnecessary

[@zrmsraggot]

It's a featuring

[@bahadr4331]

Signal processing is not a standalone job title except some very intense engineering fields which are very rare. At the end of the day, you have to earn money. If there is a simple way to do it, it will always win. But it is a must to learn intermediate-level signal processing but never lay on it alone.

[@torrance1175]

Why not?

[@bekirbakar]

I agree with you Bahadır.

[@mmarques2736]

I disagree, it is my job title (working in wearable devices), and the job title of many mathematicians in industry jobs these days... Not that rare in hi-tech industry!

[@bassman9261995]

A quick google search of “signal processing jobs” may be enlightening

[@zrmsraggot]

Sparsity, dimensionality reduction and Interprtability