Abstract: In a digital communications system, one of two tones is selected for each bit, a first tone for a "0" and a second tone for a "1"; the tones are unique for each bit. The tones for all of the bits of a byte are sent simultaneously, in a burst or pulse. At the receiver, a parametric model-based spectrum analysis is performed to find the constituent frequencies which produced the composite spectrum. This spectrum analysis uses autoregressive parameter estimation techniques to determine the most likely combination of frequency components which would have generated the composite spectrum. The combination of frequency components so found is then decoded to yield the corresponding bit values represented by those frequencies. The use of autoregressive spectral analysis techniques to recover the frequency components in the composite signal provides both bandwidth compression (by virtue of increased resolution) and faster signalling rates in transmitting just such signals.