Noise cancelling inside regular rooms

- F
- Florian Mayer
  
  Contact options for registered users
posted
18 years ago

Wed, Jun 15, 2005 8:53 AM

Hi,

as a computer scientist, I'm searching for an application of sensor-networks. An interesting one seems to be noise cancelling with a set of distributed speakers and microphones, interconnected by processing nodes. I'm not very keen on dynamics and active vibration dumping, but I have in mind that it should be possible to identify maybe the loudest source of noise by triangulating the timedelays between the microphones, and then to reduce that noise by actively generating interference with the speaker-network.

What do you think? Is this possible? Easy? Hard? Impossible?

Is there literature on how to do this? Are there existing systems, maybe a car-stereo, ...?

Thanks a lot, Florian

- B
- bruce varley
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Wed, Jun 15, 2005 10:23 AM

In theory it's easy, just apply the inverse of the pressure field, and the result is.... silence. I'll offer a prediction here, that an effective system for an 'open' scenario such as you describe won't emerge for a long time if ever. You might be able to create a fair degree of annulment over a small region, if the geometry is controlled, but more than that is likely to be infeasible. You also need to remember that hearing is logarithmic, so a

99% reduction in sound power, which is likely to be very difficult to obtain, only represents a 20dB reduction, still quite audible, and probably producing a distorted and probably unpeasant variant of the original noise.

Having said that, I'm sure that there's heaps of scope in investigating the situation, just don't have too high an expectation of achieving that specific objective.

I'm reminded of the copious adverts some time ago in airline and credit card mags for speech analysis systems that could detect 'lies'. Things ain't that simple.

- S
- Scott Seidman
  
  Contact options for registered users
Vote on answer
posted
18 years ago

Wed, Jun 15, 2005 12:35 PM

"bruce varley" wrote in news:42b00136$1 @quokka.wn.com.au:

I think you'd need to model in a fairly exact representation of each room the system is run in, to deal with echoes and the like. Also, you'd need at least three microphones to deal with inherent ambiguities and the 3D nature of the problem.

Scott