In general, the first resort in these cases is some decoupling cap.
1. As DimitriP says, a few caps (at least a 0.1 uF) from the power rail
to ground is a good start.
2. I'd also add a 0.1 from the "hot" terminal of the mic to ground.
2a. It is worth looking at the OUTPUT of the circuit as well -- the impulse
noise could be coming in on any wires. Perhaps a little decoupling
on the output terminal wouldn't be all that bad. 0.01 uF from Q2's collector
to ground at 20kHz is about 800 ohms reactive. That might be too much of a load
or present too much of a roll-off for you, but this adventure might
require some compromise....
2b. The voltage divider providing the base bias is really kind-of soft --
the base is looking back into 151Kohms for its bias supply. A 0.01 uF
cap at a minimum would bypass some of the RF noise. You might also
consider reducing the bias resistors by a factor of 10 or more. It may not
be important in this circuit, but those two bias resistors are contributing
about 2.5 uV of noise in a 10kHz bandwidth. (Vn = sqrt(kTBR) )
3. Lots of folks like ferrite beads. Sometimes they help, sometimes they
don't. The idea is to add some inductance in the paths that might be
contributing interference. I've used "common mode" chokes at times.
Some of it is like throwing salt over your shoulder -- it can't hurt.
(This is a reasonable approach for "one-offs" but if you are designing
a product, more diligence is appropriate.)
If you want to try the ferrite bead idea, slip one or two over the +12 line.
If you have a toroid, wind the mic cable through it a few times. If you have
a second toroid, wind the 12V line and the ground line through it a few
times. I don't give this a big probability of curing the problem, but if all
else fails...
I'm not really well plugged in to this audio stuff, but do you really need
200x voltage gain here? (I'm thumb estimating this based on the
ratio of collector resistance and bypassed emitter resistance, ignoring
the output loading. What kind of ouput are you driving? )