|
Wednesday, March 21, 2007 Computer Speakers for Your Ears OnlyMicrosoft researchers are developing an algorithm that would allow speakers to work like virtual headphones--even as you walk around your office. By Kate Greene
More and more people are using their computers for voice communication, such as Skype and audio instant messaging. For the most part, however, using these features requires one either to be tethered to her computer by a headset or to speak directly into a microphone and keep the speaker volume low, especially in shared office space. In light of that problem, researchers at Microsoft are trying to make audio output more sophisticated. A team, led by Ivan Tashev, a software architect at Microsoft, recently began work on an algorithm that, in theory, will be able to direct sound from a set of speakers--ideally embedded in a computer monitor--into a person's ears, effectively creating virtual headphones; just a few inches outside the focal point of the sound waves, the volume dramatically fades away. Crucially, says Tashev, his algorithm could be used by a wide range of inexpensive speakers that could be put into computer monitors. The goal, he says, is to "target focused sound so that a person can walk around an office and hear" while on a video- or computer-aided audio conference call. Information about a person's location could be collected by hardware peripherals and fed back into the speaker software, allowing the virtual headphones to move with the user in real time. For example, Tashev says, a camera, either mounted on or embedded in a computer monitor, and image-processing software could determine a person's position. In addition, an array of four or more microphones on or near a computer monitor could be programmed to localize sound by measuring the subtle time differences among when sound arrives at each speaker in the array. In fact, Tashev's previous work has been to design such sound-localizing algorithms for the types of microphones that are commonly found in the bezel of laptop computers. Employing both a camera and a microphone can improve the accuracy and distance a person could roam while using the speakers. To be sure, the idea of focusing sound isn't new: military radar systems and common ultrasound equipment, used to image fetuses in utero and find cancerous tumors, have done this for years. The technology is called beamforming, and it is achieved when the sound waves from certain speakers in an array experience microsecond delays, explains Jiashu Chen, technical manager at Finisar Corporation, a data-communications company based in Sunnyvale, CA. The delayed sound waves combine in such a way that in some parts of space, the sound is canceled out, and in others, the sound grows louder. However, beamforming systems that direct audible sound, such as music or human voices, are more technically challenging to build than radar and ultrasound are, says Chen, because they must accommodate a wider range of frequencies; lower frequencies require different hardware and software considerations than higher frequencies do. Signal-processing technology has improved to the point that some commercial products use beamforming. Yamaha, for example, sells speakers for home entertainment that bounce focused sound off walls to create virtual speakers behind a listener's head. But such systems are still rare, and always pricey. |
Audio »
Share »
Digg this
Add to del.icio.us
Add to Reddit
Add to Facebook
Slashdot It!
Stumble It!
Add to Newsvine
Add to Connotea
Add to CiteUlike
Add to Furl
Googlize this
Add to Rojo
Add to MyWeb
Favorite
Print
E-mail
An Algorithm That Makes Voices Clearer
11/28/2007
Comments
nekote on 03/21/2007 at 6:23 AM
114
Through a cell phone intermediate, or better yet, a Wi-Fi enabled cell phone / headset?
Some sort of wireless "mesh" network?
This seems like the hard way, rather than the easy way.
joncolinleonard on 03/21/2007 at 7:56 AM
3
mmatyas on 03/21/2007 at 9:43 AM
1
riposte on 03/21/2007 at 9:57 AM
1
hankejh on 03/21/2007 at 10:56 AM
16
When engineers focus all of their efforts merely on what is easy to accomplish, genuine innovation stagnates (and they probably should lose their jobs). Microsoft will not get from keyboard-mouse to more sophisticated human/machine interfaces by employing a “simple and easy” R&D effort. As riposte points out above, hardware implementation of sound beaming is already an industry, but remains too expensive for home/office implementations. By mimicking this in software, Tashev adds to Microsoft’s ip portfolio and further eliminates their dependency on specialized hardware.
Thanks for the article, Kate.
aymeric on 03/21/2007 at 2:20 PM
15
hankejh on 03/22/2007 at 12:06 PM
16
Well that makes two of us! I am an ardent fan of minimalist interfaces, more likely to keep my phone off and email closed whenever I need to focus my attention on work at hand. In Tashev’s R&D, I believe Microsoft is seeking a higher purpose – a richer, more natural human/machine interface – something akin to a person-to-person interaction with less keyboarding, more eye/face/hand gesturing and interactive audio/speech recognition.
abcarterjr on 03/21/2007 at 1:49 PM
45
powered tuned propane cannons formed into
an airborne organ to snuff forest fire flames.
After the forest critters flee the flame front
of course. Organist must be certified.
blackgenie13 on 03/21/2007 at 2:17 PM
1
All I can see is that this is a good application for listening to musics within the office or so.
bug_me_not on 03/22/2007 at 7:47 AM
6
An acoustic array must be orders of magnitude larger than their wavelengths, in order to create sufficient directivity. Relying on pure interference to create local regions of higher volume just isn't going to work.
roy1 on 03/22/2007 at 9:38 AM
2
hankejh on 03/22/2007 at 11:04 AM
16
It is true that Ltd1/Yamaha have developed an overpriced ($700+), oversized (3-to-4 foot long) home theater speaker that bounces sound off of the walls of a shoebox apartment to poorly mimic a far less expensive ($200) genuine surround sound system that occupies less than half as much space.
How does that translate to Microsoft/Tashev's clearly stated goal: a widely available, inexpensive solution small enough to be integrated into a monitor bezel that directs sound to a listener in front of the computer (not bouncing throughout the entire room)?
These are totally unrelated and Tashev's current work is clearly not available from 1Ltd or Yamaha. Moreover, why would it not make far more sense for Microsoft to simply license this technology from 1Ltd (or buy 1Ltd) if their technology actually fulfilled Microsoft's objective?
roy1 on 03/22/2007 at 12:28 PM
2
http://www.eedigest.com/mnfs/Micronas-and-1-Ltd-unveil-Digital-Sound-Projector-turnkey-solution/
hankejh on 03/22/2007 at 1:03 PM
16
The article linked in your reply, while interesting, hardly opens up the possibility of a price drop from over $700 down to the consumer price point for a computer accessory feature. If anything, it demonstrates that this technology (as Microsoft envisions it) does not exist today.
bug_me_not on 03/23/2007 at 8:35 AM
6
But anyone who has done a proper study of phased arrays knows that beamforming performance ultimately depends primarily on the physical extent of an array, and little else. No amount of algorithms will possibly overcome this.
As happens often in magazines, an idea is magnified, distorted, and hyped to an eager audience, without regard for truth. Without published data, this stuff is not to be believed.