🚀 go-pugleaf

An idea for a device which would be very useful, but for
some reason doesn't seen to be available:

In sound reinforcement work, especially indoors, feedback is 
a problem.  Certain frequencies tend to be "hotter" than others
in a given room.  Usually equalizationis used to cut these
frequencies (actually wavelengths) before a show, during
set-up.  But because of the horrible effect of feedback on an
audience and change in room acoustics with an audience, it
is normal to leave large margins in the system gain to
be sure feedback doesn't happen. 

Wouldn't it be possible to use DSP to eliminate feedback 
in real time before it is noticed?  The processor would 
"look" at the spectral contents of the program and watch
for any "runaway" frequency bands, where the energy is 
increasing exponentially (??).  It could also check for 
ringing.  It might test the room before a show to find 
frequencies to watch more closely.  If it was really slick,
it would distinguish between input channels, realizing that
feedback would only come through a mic, while a synth
might make a sound which "looks" like feedback.  

Once it found feedback starting, it could reduce the gain
at the exact frequency by just enough to stop it.  Thus,
the sound reinforcement system would continuosly adapt to
changing conditions, and much more gain could be squeezed out
of a system because feedback margins could be eliminated.

So . . . Does such a thing exist?  If not why not?
It seems  1) possible  2) useful  3) saleable
Am I missing something?

Randy Brumabugh
rando@skipper.dfrf.nasa.gov