Un-Distortion Plug-In? Is this possible... yet?

There are harmonic-tracking equalizers that can, to a limited extent, remove harmonic distortion.

The only thing I have in my toolkit remotely like that is GlissEQ, which has a harmonic-series filter type but it's static. I have used it to reduce harmonic distortion from resonances, but to un-distort a guitar would require pulling apart the fundamental frequencies like Melodyne DNA, calculating the harmonic series and applying narrow filters. Complicated, but not a huge leap from where we are today. I can certainly envision such a thing a few years down the road.

In fact I've been making such predictions for a while and wondering why they haven't come to fruition yet. Predictions of super-smart software that will allow us to diagnose any audio problem and automatically fix it. I see no reason why such things can't be done.

Audio-specific DSP technology is still in its infancy and stuck in a conventional paradigm, waiting for other technological disciplines to be integrated such as pattern recognition and artificial intelligence. That will happen when either a) the market grows big enough to attract serious research, or b) we can identify a war-on-terror angle to pursue for government funding.

Hi BF...

Yes... To me it's much like nanobot tech.... "Simply" learning to deal with things on a smaller scale gives awesome new abilities.

War on terror...? Hmmm... First thought? Maybe we could find a way that it might "un-distort" terrorist thinking? ;-)

Truth tho, I expect such a device (the musical one that is...) relatively soon... I was actually hoping someone had one out that didn't work flawlessly... yet... Just to enjoy the tech advances... If there's money in it someone will chase it.

Hi Craig...

Yes... It might be another version of the un-terrorist version created for military use...

Keni

bitflipper


The same group that nevertheless buys into all kinds of dubious voodoo such as magic pebbles.

I just got suckered into buying some of those. It was on my birthday too! Felt like a fool but at least I never bought into the war on terror... in fact I sold both my vehicles in protest. Haven't bought a drop of gas or oil in three years. Not an easy decision but I'm not sorry I made it. And if Neil Young hadn't burned down that warehouse in San Fran, I'd be sittin' behind the wheel of an electric right now.... oh well, thanks for trying, Neil! :)

The problem here is that what we call 'distortion' is based on a preconception of a known 'undistorted' tone. But to the computer it's just a signal the same as any other, and it can't know for sure what is the added component you want removing and what is the bit you want to keep.

It's fairly trivial to pick out the basic tone in a distorted guitar sound, and remove the frequencies that don't lie on the fundamental or the predictable harmonics. But some of that will have been the unique character of the pickups, the strings, even the wood of the guitar. A signal processor has no way of knowing how much of that distortion from the ideal waveform came from the effect you're trying to remove and how much came from the charactistic timbre of the instrument.

One way of approaching this would be to give it something to aim for. If you could give it a sample of the same instrument without distortion, it can attempt to leave in those characteristics while removing everything else. It still couldn't be perfect, because of information that is completely lost in the distortion process as well as in the digitisation process, but it could get close. It would probably be good enough to re-amp the signal through a different VST, which is probably the biggest commercial application for this.

Comparisons with image processing are interesting, because they use many of the same types of algorithm we typically use on audio. However, although we see a lot of impressive image enhancement on TV, most of that stuff still just doesn't exist in the real world, because you can't create information out of nothing. What you often get instead is that the computer might make several different clean-up versions of an image based on different algorithms (eg. low pass/high pass filters, pseudo-colouring, etc) , and then a human has to bring in his or her external knowledge to see if any of the new versions match something they know to exist in reality.

I continue to be impressed with what Izotopes Ozone RXrestoration program is capable of (we got the less expensive version last fall).  It, indeed, does have a "remove distortion" function and the demo of it is quite remarkable...haven't had to try it yet tho.  Head over to their site Keni, and give it a listen.
 
http://www.izotope.com/products/audio/rx/