I've actually built the thing. As to its operation, first consider one antenna. If you move it towards a fixed transmitter, the frequency seen by it will increase due to the doppler effect. If moved away, the frequency seen will decrease. Now imagine this antenna on the edge of a wheel which rotates. This will put a sinewave frequency modulation on a received carrier, and a tone will be heard from an fm receiver . We don't want to carry a big wheel spinning at a fair rate, so we simply switch between two antennae. If the antennae are at different distances from the transmitter a square wave (in theory) will appear at the audio output of the receiver. A synchronous detector, driven from the same signal used to switch the antennae, will provide a dc level dependent on the phase of the audio, which can be used to light leds to indicate which antenna is nearest. You look for the audio null to indicate direction, (both antennae at he same distance) , and rotate the thing a bit to determine which antenna gets nearest ( is it in front or behind).
You will be listening for the disappearance of a specific tone, so there will be no trouble with speech on the signal. Since it isn't synchronous, speech will average to zero at the detector and will not affect the leds.
For 144Mhz (2 metres) the antennae are dipoles , each made from 2 "rubber ducks", mounted at the ends of a bit of wood about 70 cm long.
For 450Mhz or so, the spacing between the antennae should be 25-30 cm.
Be sure that the cables from the antennae to the switch are EXACTLY the same length, or the thing will squint sideways.
One day I'll figure out how to send circuit diagrams, but until then, the active components are as follows:
2 of dual gate mosfets( antenna switch)
4013 ( oscillator, divider)
4066( synchronous detector)
lm324( amplifier, led driver)
I may still have missed something, it was a few years ago. Maybe I'll have to build another one.