Generalized Modulation and Iterative Demodulation

Modulation with correlated signal functions is considered. While it is well-known that for low spectral efficiencies linear separation via minimum mean-square error (mmse) filtering provides close to optimal performance, linear approaches fail as higher spectral utilization is desired. The alternatives of approximations to maximum-likelihood decoders, such as sphere decoding quickly become practically infeasible due to complexity constraints. In this paper an alternate method is proposed where the redundancy required for error control is achieved by increasing the number of signal functions used. It is shown that the capacity of the increased set of signal waveforms is nondecreasing and achieves the capacity of the Gaussian multiple access channel as its upper limit, when the number of waveforms becomes large. Furthermore, a simple iterative demodulator allows achievable spectral efficiencies beyond those available by linear processing, and it is proven that the capacity of the channel can be approached to within less than 1 bit as the number of signal functions becomes large.