"When we asked him to provide a simple explanation of what happens, computationally, when a voice signal enters his software, he opened his desk and pulled out thick stacks of folders, each stuffed with hundreds of pages of mathematical equations."
Perhaps, but your comment doesn't teach us anything. It would be better to assume a smart but ignorant reader and walk them through specifically how it's bullshit.
It's journalistic embellishment. You would need the equivalent knowledge of a bachelor's degree in EE to understand the algorithm, but it's really just that he strung together common primitives in signal processing like a resampler and an auto-correlator.
It's quite clever really, suggesting so much while not actually saying anything at all. Imagine someone pulling out their undergraduate notes to show you some basic equations, and describe it in the most florid way possible - you get something like the quoted sentence.
Did anyone else notice that in the figure with caption "A rendering of the Auto-Tune interface; via WikiHow", there's a lynda.com watermark in the lower bottom right corner of the screen? The screenshot used for the illustration must've been taken from a lynda.com video course.
It's funny how we all know this type of technology exists, but we never stop and find out the backstory of why and how it came to be. Great article, but depressing in that I don't understand 50% of the content and am unlikely to ever do so, despite having two friends in the geo business - one collecting the data and one interpreting it.
8 comments
[ 3.2 ms ] story [ 26.7 ms ] threadThis is utter nonsense. The math in the patent would fit in a Tweet. https://patents.google.com/patent/US5973252A/en
I've been reading weirdo articles about this technology for 15 years. Here's a good summary of where the coverage goes wrong:
https://valhalladsp.com/2009/05/21/auto-tune-autocorrelation...