IDMf - Electronic Music & Production Forum - IDMforums.com - View Single Post - "Compressive Sampling" (Compressive Sensing)...WTF????!
View Single Post
Old 03-05-2016, 08:22 AM   #2
Blingley
Madacu Sad Gent
Blingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MCBlingley is a savage MC
Blingley's Avatar
Posts: 1,867
MC Status: 7238579
Thanks: 248
Thanked 837 Times in 556 Posts
Re: "Compressive Sampling" (Compressive Sensing)...WTF????!

It really is not that complicated. If we can assume that the signal contains less data than the maximum amount of possible data (In other words: The signal is not pure noise), we can reconstruct the signal from an according number of samples. In other words, because signals are not noise and they contain redundant information, we can reduce them into domains where there are a lot of zero coefficients. Think about your basic piano chord for example. If you just look at the samples themselves, or an oscilloscope shot, it may not seem particularly sparse as a signal. If you look at a spectrograph instead, you will notice that there are a lot of frequencies that are not present in the signal. Essentially the signal is sparse in the frequency domain, but not sparse in the time domain.

Thus we can sample something, say for example an image matrix or a musical excerpt, and we do not need particularly many samples for this. We will essentially get parts of equations from here - or we get values that need to conform to the equations behind them. No unique solution necessarily exists, as we have more unknowns than functions, but if we can assume that the signals are sparse in some domain like the piano chord example before, we can find a satisfactory solution.

------------------
Blingley is online now   Thanks Reply With Quote
Thanks to Blingley
Nystagmus (04-05-2016)