Multiuser decision-feedback detection: performance bounds and adaptive algorithms

Multiuser decision-feedback detection has been previously considered for DS-CDMA. Here we evaluate asymptotic MSE and capacity (averaged over the users) of the multiuser decision-feedback detector (DFD) in the absence of error propagation. Asymptotic means that the number of users K/spl rarr//spl infin/ and the processing gain N/spl rarr//spl infin/ with K/N fixed. The resulting capacity is the same as that achieved with optimal detection. (This is also true for finite K and N.) We also present and compare adaptive stochastic gradient and least squares algorithms for updating the DFD coefficients without explicit knowledge of channel state information.