Bayesian turbo multiuser detection for nonlinearly modulated CDMA

This study introduces a blind iterative multiuser detection scheme for demodulating nonlinearly modulated CDMA signals in Gaussian and non-Gaussian noise, based on the Bayesian inference. The detector jointly estimates the channel gains, the noise parameters and the transmitted symbols. It yields asymptotically to the MAP receiver (or minimum error probability on decisions) by using the Gibbs sampler, a Markov Chain Monte Carlo method for Bayesian computation. Its performance is compared with that of the coherent parallel interference cancelation receiver and an iterative detector based on interference suppression. It is seen that the proposed Bayesian multiuser detector offers substantial performance gain over those existing methods, especially in a highly loaded system and in the presence of non-Gaussian ambient noise. Being a "soft-input soft-output" algorithm, the Bayesian multiuser detector is particularly well suited for turbo processing in a coded system: the multiuser detector can incorporate the a priori symbol probabilities, and it produces as output the a posteriori symbol probabilities; the channel decoder delivers information about the code bits prior distributions, which can then be fed back to the multiuser detector to refine its processing, and vice versa. Hence, the proposed Bayesian multiuser detector naturally leads to a turbo receiver structure for joint multiuser detection and decoding.