Noise begone! New tech banishes camera noise from video soundtracks

by Dave Etchells

posted Wednesday, May 23, 2012 at 1:07 PM EDT

Silicon Valley Enterprises' logo. Click here to visit the Silicon Valley Enterprises website!You heard it here first: Zoom and AF-motor noise in video soundtracks could soon become a thing of the past.

A real limitation of movie recording with digital cameras is that motor noise from zoom and autofocus systems is glaringly obvious in video soundtracks. Dedicated camcorders use special motors and sound-deadening in the camera body to eliminate this problem, but they typically have more space to play with than's available to digicam makers, so such measures are easier to implement. When it comes to cameras with interchangeable lenses, many lenses you might want to use weren't designed for silent focusing, so you're more or less stuck unless you shell out for special video-specific lenses.

So what to do? Some manufacturers simply don't let you zoom while you're recording a movie, while others allow zooming (perhaps slowed down to reduce noise), and leave it to the user to decide whether having motor noise in their sound tracks is worth it or not. (Personally, I'd say it almost never is.)

It's no secret that video-capable digicams and SLRs / compact system cameras are pushing dedicated camcorders out of the market, but the motor noise issue is a serious limitation. Cameras without the problem would have a significant leg up on their competition in a highly competitive market.

Example of a modern digital camera image processor. SVE's VideoZizzleâ„¢ is designed for easy implementation in camera designs without changes to motor systems or signal processing hardware. Image provided by Canon.

Example of a modern digital camera image processor. SVE's VideoZizzle™ is designed for easy implementation in camera designs without changes to motor systems or signal processing hardware.

As it happens, new technology promises to virtually eliminate the motor noise problem. Better yet, it can be dropped into most digicam or SLR / CSC designs with no change to either motor systems or signal processing hardware.

The company behind this new tech is Silicon Valley Enterprises (SVE for short), a group of clever engineers specializing in digital signal processing of audio signals. They call their new solution VideoZizzle. (They're clever at DSP engineering; product naming perhaps not so much.) ;-)

The idea behind VideoZizzle is that, with the right signal-processing algorithm, the camera's processor can recognize and surgically remove the sound of the zoom or AF motor, leaving the rest of the audio relatively untouched. Their algorithm also appears to be adaptive, requiring no advance knowledge of what the motor noise sounds like. As a result, you can pretty much drop the VideoZizzle code into an arbitrary camera and it'll recognize and remove motor noises from the audio track in real time, while the video is being recorded. As noted, all using the camera's existing processor.


Before-and-after comparison of an audio waveform processed with SVE's VideoZizzle software to remove the lens zoom motor drive noise.
Screenshots provided by Silicon Valley Enterprises Inc.

I have to confess that I had pretty modest, even negative, expectations for the product when I first heard of it: I have some experience in sound engineering, enough to know that attempts to remove specific sounds from an audio track almost always have unfortunate consequences for the part of the audio you're actually interested in.

I was pretty well dumbfounded, then, when I listened to the examples of the Zizzle technology on their website. (Actually, dumbfounded doesn't quite describe the feeling: My Brit friends use a term, "gobsmacked;" that better expresses the magnitude of it.) The video examples on SVE's website are rather prosaic, the sort of things engineers would come up with to demonstrate something. (I can say this without pejoration, being an MSEE myself.) But by golly, they demonstrate the effectiveness of Zizzling. Check out the examples below; the first segment in each is the original audio, and  the second segment is the audio with VideoZizzle applied.

First, a basic demo with modest background noise and a human voice:

Next, an example over music ("the sound of Carlos Santana, ladies & gents"), where you might expect to hear more distortion, given the broader range of audio freqencies involved:

The effect isn't 100%: If you turn up the playback volume enough, you can still hear the faintest whisper of the zoom noise in the background. But that's frankly picking at nits: for all intents and purposes, the zoom noise is just gone.

Silicon Valley Enterprises' play here is to license the VideoZizzle code to camera manufacturers to incorporate it into their digicams, interchangeable-lens cameras, and even camcorders. (Spend a little on firmware, maybe cut out a bigger chunk of cost in the form of expensive, silent lens motors.) So whether we'll actually see this in the next generation of cameras will depend on a range of factors, including SVE's sales ability, how they price their product, and the openness of camera makers to the whole idea.

Personally, unless the licensing fees are truly prohibitive, I think camera makers would be crazy not to want to incorporate VideoZizzle in their products. More than most camera technology I've seen in the past 15 years or so, it delivers really compelling end-user benefits. Stated plainly, given the choice beween two identical cameras, one having VideoZizzle inside and the other without, most consumers would opt for the Zizzlized version in a heartbeat. Manufacturers should adopt this and actively tout it to consumers as a key feature.

On their side of the transaction, SVE would be crazy to charge too high a price for it: Priced right, there's no reason they shouldn't expect to see their tech inside every digicam on the market.

Here's hoping that both SVE and the manufacturers have their heads on right in approaching these partnerships.

At this point, it's by no means a done deal or foregone conclusion that we'll see Zizzle-equipped cameras soon if at all. I'm rooting for SVE, though, and feel pretty confident in predicting that a couple of years from now*, zoom motor noise in videos will be strictly optional. A few years after that, and it'll be a thing of the past. For myself and a few hundred million other camera users around the world, that day can't come too soon.

* I say a couple of years from now, because digital cameras typically have about a 18-month development cycle. Given that this is firmware-only, though, it's possible we could start seeing it a lot earlier than that (maybe announcements as early as next year's CES?), as it could be dropped in later in the development cycle than something requiring new hardware to implement.