Markov chain Monte Carlo (MCMC) methods are considered as the powerful techniques for Bayesian computation.However

it is difficult to apply them to the asynchronous CDMA systems in the presence of unknown interferences and multipath fading.In this paper

a novel Bayesian multiuser detection method is proposed to overcome the difficulty.In the proposed method

the received signal is preprocessed at first by the linear group-blind decorrelator

and then

the Gibbs sampler (a typical MCMC procedure) is employed to perform the Bayesian multiuser detection.Simulation results show that this method significantly outperforms the linear group-blind multiuser detection with a low additional complexity linear with the number of intra-cell users.To improve the performance of the proposed method

two stage Gibbs samplers are used.In the first stage

the parameter estimation with high accuracy is achieved based on the output of the Gibbs sampler and the estimated parameters are used in the Gibbs sampler of the second stage.It is seen that the detection performance is improved in comparison with that of the method using one stage Gibbs sampler.