IIRC, and that may not be the case, the way we did it back in the late '70's with our rather slow 8-bit micro's was to differentiate the signal coming to the receiver, and time the distance between the pulses. A pulse occurred at the positive and negative peaks, but, AC coupling (differentiating) the signal and biasing it to GND meant that one only got a positive spike when there was a positive edge, i.e. the slope of the incoming waveform had switched from negative to positive. This enables one to discriminate between the high and low frequency. Some reasonable wave shaping has to be done at the transmitter.

Keep in mind, too, that this was done for use with low-cost audio-cassette recorders with limited frequency response. One guy used a pretty good tape recorder and found that the higher quality actually made things more difficult.

RE