7 Conclusions

This new analytic technique is capable of excellent separation in a low noise environment. This has been verified by tests of both simulated and field data. An extension to $M\ge 2$ signals is straightforward, with the use of a trellis with $2^{M-1}$ states. It appears that in order to successfully demodulate a subdominant signal, about a 20dB power separation must exist between the subdominant signal and the noise. Further refinements of the algorithm may make it more resistant to noise.

Acknowledgement: The author thanks an anonymous reviewer for pointing out the need for AWGN bias corrections.