Studies On Algorithms For Some Problems In The Field Of Communication Signal Processing

This dissertation is devoted to studying the algorithms for blind equalization, multi-user detection and blind multi-user detection problems in the field of communication signal processing. They are all used to eliminate the interference of communication systems. The blind equalization is to remove the inter-symbol interference, and the multi-user and blind multi-user detections are to cancel multiple access interference. Main contents of this dissertation are as follows:In chapter 1, the background and motivation of this dissertation are introduced and some existing methods for blind equalization, multi-user and blind multi-user detection are surveyed. Finally, main research work of this dissertation is discussed.The blind equalization is studied in chapter 2. To resolve the problem that the existing blind equalization algorithms are hard to be applied in high-speed communication systems, a new blind equalizer, named as algebraic equalizer, is proposed. It only uses operations for comparison and few multiplication and addition such that the computational efforts are much lower than those of usual blind equalization algorithms. Another difficulty in designing blind equalizer is how to solve the problem of non-minimum phase since most of existing methods could only solve the minimum phase parameters of MA model. Based on the model of inter-symbol interference, the relationship between the blind equalization and parameter estimation of MA model is deduced, and the relationship between parameter estimation of MA model and signal reconstruction from its spectral magnitude is established. As such, all the theory and algorithms in signal reconstruction from its spectral magnitude should be applied to solving the parameters of MA model.The multi-user detector is studied in chapter 3. Five algorithms of multi-user detectors are proposed. The first four of them are based on the optimum multi-user detector that is modeled as a binary quadratic programming. The fifth finds the relationship between nonlinear equations and multistage multi-user detector, and thenew detector improves multistage detector. Our simulation experiments show that the bit-error-ratio (BER) of these new detectors is much lower than that of decorrelation, minimum mean-squared error and multistage detectors. Finally, we summarize and compare their respective features.The blind multi-user detector is studied in chapter 4. Based on minimum mean-squared error criterion, the existing blind multi-user detection (SBMUD) is relies on subspace decomposition algorithm. In this chapter, the blind multi-user detection is described as the problem of solving ill-conditioned linear equations. It is shown in this chapter that SBMUD can be also obtained by the truncated singular decomposition method used to solve ill-conditioned equations. Thus it can be concluded that the essential problem of blind multi-user detection is how to solve ill-conditioned equations. The significance of this link is that by considering the problem of blind multi-user detection as that of solving ill-conditioned equations of mathematics, many mathematical methods, like regularization methods, can also be used to solve this problem. Furthermore, a new blind multi-user detector based on Tikhonov regularization is proposed together with a strategy for estimating the regularization factor. Moreover, a new criterion, called min-max output energy criterion, is proposed. Based on it, a new blind multi-user detector is proposed in this chapter. It is different from the existing blind detector in that regardless of whatever the statistical properties of signals are, the new detector always obtain the satisfactory when there are no channel noises.In chapter 5, the author concludes this thesis and gives prospects of the further research in the fields.