Researches On Blind Parameter Estimation In MIMO-STBC Systems

In recent years, the development of multimedia and Internet make rigorous demands on the reliability and rate of wireless communication systems. Multiple Input Multiple Output(MIMO), which can efficiently promote the channel capability, is one of the most effective solutions of these problems. The transmitted signals are introduced redundancy to obtain diversity gain by Space-Time Coding(STC) in MIMO systems. The Space-Time Block Code(STBC) is the major category of STC. MIMO-STBC systems, one of the key techniques for the next generation of wireless communication systems, have been standardized by IEEE 802.11 n and IEEE 802.16 e.Blindly estimating the parameters of communication systems is an important research field with military and civil applications. The cooperative MIMO-STBC systems have been investigated vastly. However, the blind estimation of parameters in non-cooperative MIMO-STBC systems has been scarcely reported in literature. Independent Component Analysis(ICA), one of the most important signal processing methods, has been developed in 1990 s. In contrast with other methods, ICA is one of the most powerful tools for the non-cooperative signal processing since it does not need strong a priori assumptions. This thesis studies the blind estimation of the parameters in non-cooperative MIMO-STBC systems by adopting ICA as a mathematical tool. The main results of this thesis are:1、A source number estimation algorithm suitable for spatially colored noise is proposed. The number of signal sources is an important parameter in MIMO-STBC systems. Most of the methods in literature assume that the noise is spatially uncorrelated. However, this assumption can not hold for MIMO channels subject to interference, and these methods can not work in this scenario. In this thesis, the noise subspace is estimated by eigen-decomposing the cumulant matrices of the intercepted data. After unitarily transforming the received data, the separated data corresponding to the noise subspace follows Gaussian law. To estimate the source numbers, the kurtosis is adopted as a statistics to test the Gaussianity of the separated signals. The new algorithm is applicable to spatially correlated noise with good performance of correct detection.2、The question of recognizing STBCs in a high dimensional feature space is studied. The cyclostationarity is utilized as feature to recognize the STBCs. In this thesis, the autocoorelation matrices with different time lags are computed. The Frobenius norms of these autocorrelation matrices have peaks and valleys at different time lags for various STBCs. Therefore, the norms can be adopted as features to identify the STBCs. The Frombinius norms of the autocorrelation matrices are mapped into a high-dimensional feature space to classify the STBCs. Compared with the methods in literature, the new methods achieves better performances since more information of STBCs are utilized.3、A rotating blind separation method for MIMO-STBC systems is proposed. The intercepted data can be separated by estimating the channel matrix in cooperative MIMO-STBC systems where the coding information is available. However, in non-cooperative MIMO-STBC systems, where both the Channel State Information(CSI) and coding matrices are unavailable, blindly separating the received data is a challenging problem. In this thesis, a rotating MIMO-STBC model is proposed by merging the coding and channel matrices as a Virtual Channel Matrix(VCM). The new model maximizes the independence of the source signal by rotating the received data with specific angles to meet the assumption of ICA. The new method does not utilize the coding information without precoder.4、The application of Multidimensional Independent Component Analysis(MICA) in non-cooperative MIMO-STBC systems is discussed, and a new MICA algorithm based on Higher-Order Statistics(HOS) is proposed. In this thesis, the MIMO-STBC systems are expressed by ICA model to separate the intercepted data. However, the ICA assumption can not be satisfied since the source signals are group-wise independent. MICA algorithms are adopted in this case. A new MICA algorithm is proposed in this thesis by Jointly Block-Diagonalizing(JBD) the cumulant matrices to estimate the mixing matrix. To overcome the drawback of poor convergence of the 1-step algorithm, it is proposed to solve the optimization problem of JBD by two steps in this thesis:(1) Jointly Diagonalizing(JD) the cumulant matrices;(2) removing the additional permutation ambiguity. To remove the indeterminacy, the object function of JBD is expressed in the form of cross-cumulants to show that the problem of grouping is equivalent to maximizing the cross-cumulant of sources in a group. The convergence of new method is guaranteed without setting an objective threshold.5、A blind modulation recognition algorithm based on Maximum Likelihood(ML) is proposed. Modulation recognition is an important research issue of estimating the communication parameters. However the methods suitable for non-cooperative MIMO-STBC systems have been scarcely reported. This thesis focuses on this problem by following steps:(1) the MIMO-STBC systems are expressed by ICA model on which a classifier based on ML is proposed;(2) the modulations are classified into two classes: independent and group-wise independent constellations;(3) the estimation of VCM is discussed for these two constellations respectively;(4) after partly removing the ambiguities, the classifier is proven to be insensitive to the remaining indeterminacies. The new method can work in non-cooperative MIMO-STBC systems with high probabilities of correct recognition.