Sequential Monte Carlo methods for data detection and channel parameters estimation

Mobile wireless channels are characterized by time-variation and multipath propagation which result in intersymbol interference (ISI) and sever distortion of the pulse shape of the transmitted data. Data detection in such environment requires the development of equalization techniques that can reverse, or at least minimize, the deleterious effect of the channel. The design of such equalization techniques is challenging as the equalizers are required to be sufficiently fast to adaptively estimate the time-varying channel. Moreover, prior to equalization, accurate estimation of the synchronization parameters is critical for reliable communication. Unfortunately, optimal estimators for synchronization parameters are impossible to obtain in general and, therefore, most of the existing techniques are based on approximate and heuristic methods.; In this thesis work, we developed novel equalization and synchronization algorithms for data detection over mobile wireless channels. Our approach to the problems is based on the Bayesian methodology by deriving sequential Monte Carlo, also known as particle filtering, algorithms. We regard the application of particle filtering in this context very appealing, as it is capable to yield near optimal numerical solutions when, as in this case, analytic solutions do not exist.; In the first part of the thesis work, we deal with the equalization problem. The problem is first formulated as dynamic state space (DSS) model where the channel values and the transmitted data are considered as the hidden states of the model whereas the received signal is taken as the observed signal. With this formulation, we have developed blind algorithms for joint data detection and channel estimation in frequency-selective channels based on particle filtering. The proposed algorithms are flexible and efficient and have shown, through computer simulations, better performance than the per-survival-processing which is the popular conventional method.; In the second part of the thesis work, we develop particle filtering based algorithms for the joint estimation of synchronization parameters and the detection of transmitted data. In doing so, we first represent the problem in an extended DSS model where we assume the symbol delay and the frequency offset variation to be first-order autoregressive (AR) stochastic processes. Furthermore, the flat-fading channel, whose complex coefficient can be considered as the signal amplitude attenuation and phase offset, is modeled by a second-order AR process driven by a complex white Gaussian noise. With this formulation, efficient and blind algorithms based on particle filtering methods are developed. In the development of the synchronization algorithms, two receiver configurations that remove the ISI and consequently that enable us to attain close-to-optimal symbol error rate (SER) are proposed. Computer simulations demonstrated that the developed algorithms achieve close to optimal performances and are superior to receivers based on the conventional timing error detectors (TEDs).