Parameters Estimation And Data Detection For CDMA Mobile Communications

The performance and capacity of code division multiple access (CDMA) systems are severely limited by multiple access interference (MAI), inter-symbol interference (ISI) and frequency-selective time-varying fading of mobile channels induced by multi-path effect, Doppler effect and non-orthogonal spread-spectrum codes between the active users. The new generation mobile communication systems are expected to provide to higher data rate, more reliable communication and faster mobile terminal. We research the estimation of active users, channel estimation and MUD based on Monte Carlo Bayesian idea for the objective of optimization estimation and detection in the uplink and downlink of CDMA mobile communication systems.The CRB and MCRB expressions of multi-user signal phase, time delay and amplitudes are given whether the training sequences and spread codes are known in the multi-user multi-path long code DS-CDMA systems. The effect of the number of training sequences and the parameters of active users to CRB and MCRB is simulated. The CRB and MCRB of multi-path DS-CDMA systems are compared to the single path DS-CDMA systems'and the bound of algorithm performance is presented for the parameters estimation. The impact of the channel estimation errors and active user'parameters estimation errors on MUD are analyzed respectively in the multi-user synchronous multi-path DS-CDMA systems. The impact of estimation errors of active users'parameters on linear de-correlation MUD, linear MMSE MUD and direct matrix inverse MUD is researched employing matrix theory.We introduce the Reversible Jump Markov Chain Monte Carlo (RJMCMC) method to estimate the number of active users and detect the signatures in CDMA mobile communication systems. The state space of active user and the space movement are described associated to the case of cell networks. The posterior distribution of demand parameter is achieved through integrating the nuisance parameters of posterior probability density function and the satisfied condition of signal amplitude density parameter. We analyze the problems of signature sequence detection and timing information estimation in cell uplink MUD using Rao-Blackwellised (RB) particle filter (PF) algorithm. The complexity of RBPF is lower than ML. Simulate results show the detection performance of RBPF is lower than ML in Gaussian white noise and has almost equal to the ML in the colored noise.The channel parameters of mobile communication are one of key factors that impact on the CDMA receiver performance. Exception maximum statistic approximation algorithm for the slow frequency-selective fading channel and particle maximum a posterior estimator and online exception maximum estimator algorithms for the fast frequency-selective fading channel are researched based Monte Carlo Bayesian idea through establishing channel response dynamic equation and utilizing observation equation. Two kinds of channel estimation methods based MCMC are supposed in the frequency-selective fading channel. Exception maximum algorithm estimated channel parameters with batch program in the unknown noise variance. The sampling data are all DS-CDMA symbols. The exception of exception maximum algorithm may not been computed owing to the high dimension of probability density and can been approximated through the approximation of posterior density. The unknown states and parameters are estimated sequence according to the particle filter which decreases the complexity of computation and store requirement. Two methods based on the particle filter improved the performance of parameters estimation due to introducing the parameters space diversity.The CDMA receive signal is considered as two dimensions compound Poisson process vitiated by the noise with the parameters of the number of active users and paths. The channel sparseness impulse response including amplitude and the timing information denotes the Poisson process in time axes and is modeled by the means of a Bernoulli-Gaussian process. Then, the optimization of the resulting posterior distribution resorts to Monte Carlo Markov Chain simulation techniques to implement the active users'identification and channel estimation. Simulation results show that the proposed method offers better estimating performance than expectation maximization approach and iterated conditional mode.The problems of blind turbo MUD in unknown multi-path channels for asynchronous coded CDMA systems employing long spreading sequences are addressed. A novel blind Bayesian MUD is proposed, which computes the MAP estimates of the channel coded multi-user symbols that are differentially encoded before being sent to the channel. This technique is based on the Gibbs sampler, a MCMC technique for Bayesian computation. Another issue addressed is blind Bayesian MUD in the presence of unknown out-cell multiple-access interference (OMAI), a scenario that occurs in CDMA overlay systems. We propose a blind approach to interference suppression, where the total effect of white Gaussian noise, OMAI is modeled as colored Gaussian noise with some unknown covariance matrix and the Gibbs sampler is then used to calculate the Bayesian estimates of all unknowns. Joint the identification of the active users and MUD are researched finally. The posterior probability density function of RJMCMC algorithm is shown through introducing the subspace concept. Joint the identification of the active users and MUD are implemented using two-layer-nesting iterative method and based-Gibbs sampler Bayesian estimation in DS-CDMA cell systems.Two methods including Markov Chain RB method and based-uniform sample Markov Chain RB (MCRB-U) method are proposed according to the differential prior density of samples. The performance of MCRB-U method has better than the former since the candidate samples of MCRB-U method are collected in the high probability range. Then, we analyze the convergence character of MCRB-U by the one-step transfer probability and state transfer equation. The two ways of circuit implementation of MCRB-U method, serial and parallel circuits, are presented according to implementation equation of Gibbs Sampler and log likelihood rate equation of MCRB-U. The word length of circuit implementation is simulated.