non-polarized electrolytic capacitor
Yea -- you are right, except you would see a diode drop in voltage, so the cap would only be good for unipolar applications -- what I mean is, consider what happens when you apply a large sine wave across the cap -- you only have conduction of the diodes when the signal is bigger than 2 diode drops -- it's a lot like the crossover distortion encountered with a class-B amplifier.
Added after 1 hours 16 minutes:
Actually, I was thinking about diodes in series with the caps, and then those two branches in parallel. Sorry.... But I think the diode drop issue is still valid in general - you might still run into problems if you need to couple a large signal AC (i.e. bigger than 1-2 diode drops) voltage through the cap.