Crosstalk cancellation, or “Why Dr. Edgar Choueiri is my hero” PART 2
Alright, so in Part 1 we established an understanding of what crosstalk is, and why it hinders binaural audio playback through speakers.
Here’s a little recap from Edgar Choueiri himself, with a quick demo of his crosstalk cancellation:
(Unfortunately, the demo doesn’t really work with MacBook speakers, so plug in your desktop speakers if need be)
So why then, is Choueiri my new hero? There are two reasons.
The first is technical. I can’t pretend to be a mathematical whiz by any means, which is to say that it was quite a triumph for me to get through Choueiri’s paper, “Optimal Crosstalk Cancellation for Binaural Audio with Two Loudspeakers.” In it, he goes into great detail about all of the formulas that relate to graphs of frequency response from different types of crosstalk cancellation. It’s very impressive work.
*Here are some of the key points:
- The major problem with most filters for crosstalk cancellation is that they bring with them “spectral coloration” — they boost some frequencies and adversely affect the sound colour of the recording.
- As Choueiri says, “the ‘perfect’ [crosstalk cancellation] filter would impose these necessary level boosts only at the loudspeakers in such a way that, at the listener’s ears, not only the crosstalk is cancelled, but also the frequency spectrum is reconstructed perfectly, i.e., with no spectral coloration.”
- Other developers (namely, Takeuchi and Nelson) have created crosstalk filters that avoid coloration, but require 6 speakers for playback — immediately pushing away non-audiophile listeners who don’t have a multispeaker playback system.
- One of the ways to filter crosstalk is through “constant-parameter regularization,” and while it works fairly well, it introduces significant coloration, especially in bass and high frequency response.
- A smoother way to filter crosstalk is through “frequency dependant regularization,” which applies different filter responses to different bands of frequencies. This is what Choueiri has been developing.
(*Disclaimer: I’m not a scientist — this is just a quick synopsis!)
Neat theory, right? Audio jargon aside though, here’s the second, more important reason that I think Choueiri’s work is so significant:
Smooth crosstalk cancellation using only two speakersmeans that just about anyone — from audiophiles with stereos worth three times my car, to my dear sweet grandmother with her boombox CD player — will be able to enjoy beautiful binaural audio more easily than ever before.
It’s great news for the binaurally-obsessed like me, who are constantly trying to get people to pull out their headphones and check out a new recording. While personally, I still prefer binaural recordings on headphones for their portability and isolation from outside sound, there are a lot of people out there that for whatever reason, prefer listening with speakers.
For a full audio demonstration of Choueiri’s work, check out this link. One thing I noticed though: when you listen to the speaker demo, make sure you’ve got the speakers setup only about a foot or so apart, and at about head height. You might have to fiddle with it to get it just right, but trust me, you’ll know when you do.
Once you get a chance to check it out for yourself, let us know what you think!