Iterative channel estimation for turbo receivers in DS-CDMA

The work considers the problem of channel estimation in the iterative reception of pilot-aided signals in DS-CDMA systems. The performance of classical training-based schemes is severely degraded in highly frequency selective channels due to the code-multiplexing of traffic and pilot signals. Thus, estimation algorithms that rely on the presence of the pilot signal, but also consider the information signal structure, are preferred. We present a Bayesian channel estimation algorithm for turbo receivers that effectively exploits the available soft information about the symbols (to model the traffic signal) in order to improve the channel estimation iteratively. Simulation results in realistic frequency selective test cases reveal only a moderate degradation compared to the perfect channel knowledge case.