I tried this yesterday but the results were not too promising.
My dev environment is a Macbook not connected to ground. The sensor is a 3"x3" piece of aluminum foil that I stuck under the cardboard of a hardcover book, R=10MOhm. The sensing range for my hand was around 0.5" and for a single finger
For 10" I was using a whole hand and somewhat larger sensor - it just depends on what you are trying to accomplish.
I could just make out a tiny increase of values upon touching the insulator. Touching the foil itself yielded a nice huge delta. Replacing the 10meg with 20meg resistors hardly changed the results. Moving my hand closer to the laptop however, DID create a large change in values.
This probably had to do with the wire connecting the sensor, which is live (sensing - unless you use shielded cable). The Theremin sensor though reported a similar effect.
The docs say to use a grounded connector which means that you could only build tethered projects. Kind of limiting for an embedded device such as the arduino- it would be nice if both battery operation AND power supplies would yield similar results.
The Qprox datasheets suggest taping a ground to the wall - The sensor doesn't need to be connected to a laptop - a power adapter works equally well (no need to be grounded). Unfortunately - the calibration is going to change a bit. As far as I can see there's no way to avoid this without adding an earth ground.
The approach I am using with subtracting the baseline should add some kind of rough calibration, but there's a lot I don't know about this yet. If you've used the Qprox linear sensors, you know how jittery and glitchy they are too.
The frequency approach isn't really that different - think of the whole thing as a big oscillator and you'll see that the frequency is changing. The period of the oscillation is being measured.
See the theremin approach in the playground (and posted on this forum too).