Re:The Mathematical Genius of Auto-Tune
by conistonwater on Monday October 16, 2017 @07:23PM
(#15487467)
I would like to see them dramatize Dijkstra's algorithm in the same style. Instead of looking at all possible paths of arbitrary length within the graph, of which there are infinitely many, ...
FFT is a pretty cool thing. But yeah, this "trick" has been used for correlation-based computations for quite awhile.
Assuming that it's the trick that the author was referring to, the article is actually a bit sparse on details.
Re:The Mathematical Genius of Auto-Tune
by soundwave106 on Monday October 16, 2017 @04:48PM
(#15486531)
The patent for Autotune can be found here: https://www.google.com/patents/US5973252?dq=5,973,252
My (very non-expert) reading of the patent is that it uses sample reduction and some specific features of the periodic form of the human voice to simplify the math involved in the auto-correlation routine. Although this routine does seem to be unique as far as I know, I wouldn't be surprised if it shares similarities to other techniques to reduce correlation complexity.
Re:The Mathematical Genius of Auto-Tune
by hamiltonkibbe on Monday October 16, 2017 @05:21PM
(#15486728)
Polyphase filters maybe? It's an approach used primarily in resampling, and its an easy way to skip a LOT of Multiply-by-zero-and-Accumulate-ing
Re:The Mathematical Genius of Auto-Tune
by nejenendn on Monday October 16, 2017 @09:32PM
(#15488002)
It decoded to a phase vocoder on the nearest semitones.
read the patent - all time domain, no phase vocoder involved.
> “I realized that most of the arithmetic was redundant, and could be simplified,” he says. “My simplification changed a million multiply adds into just four.”
Doesn't sound like FFT.
at the era of deep learning, no one is impressed by their 500 summations of multiply adds, or 1 million multiply adds.
A geforce 1080ti does 11.3 TFlops, with a T, that's 11 300 000 000 000 floating point operations per second..
You should still be impressed with the efficiency.
Just cos you've got enough grunt to do it the dumb way doesn't mean the smarts that produced this are somehow unimpressive.
Re:The Mathematical Genius of Auto-Tune
by jdietrich on Monday October 16, 2017 @06:14PM
(#15487083)
Audio DSP has become much more sophisticated now that we've got cycles to burn. RX and Melodyne are practically voodoo.
Re:The Mathematical Genius of Auto-Tune
by TheRealPomax on Monday October 16, 2017 @11:18PM
(#15488409)
I have no idea how RX6 does some of the stuff I make it do - all I know is that it does it better than I could have asked for.
Re:The Mathematical Genius of Auto-Tune
by smelterdemon on Monday October 16, 2017 @05:29PM
(#15486768)
In 1996 your processor might be running at 150MHz. Using a standard 44.1kHz sampling rate, that doesn't give you very many cycles to get each sample processed to do this in real time.
Re:The Mathematical Genius of Auto-Tune
by hamiltonkibbe on Monday October 16, 2017 @05:33PM
(#15486793)
That's fine, but being able to run multiple audio tracks worth of VST plugins in real time on a modest laptop is very likely more valuable to their target market
Re:The Mathematical Genius of Auto-Tune
by hamiltonkibbe on Monday October 16, 2017 @05:42PM
(#15486862)
You'd be surprised how quickly those 500 multiply-accumulates per autocorrelation point add up... a 128-tap filter * 500MAC/point * 2 channels(stereo) * 192ksps is already at 24+billion FLOPS for a single track...
CPU usage is even more of a problem that several years ago. Some very well known _compressors_ takes a whole 10% of CPU here. 10 of them and you're not realtime, probably earlier than that.
Re:The Mathematical Genius of Auto-Tune
by pocketsquare2 on Monday October 16, 2017 @05:32PM
(#15486789)
I'm impressed by a great product. What are you doing with all those FLOPs?
I think this is the article's way of dramatizing the standard way of calculating the autocorrelation using the convolution theorem: https://en.wikipedia.org/wiki/Autocorrelation#Efficient_comp...