Research On Wideband Blind Linear Equalizer

Blind equalization is one of the key technologies to compensate for channel distortion and enhance the communication quality. Since Y. Sato first introduced the idea and algorithm of blind equalization in 1975, the blind equalization technique has attracted great attentions. Many researchers have made in-depth studies on the principles and algorithms of blind equalization, and got a lot of research results. At present, blind equalization has found applications in medium-speed or low-speed broadcast communication systems, receiver of non-cooperative communication systems, and so on.In the recent years, with the rapid development of the high-speed wideband wireless communication, the date rates and transmission bandwidth of the existing single-carrier communication systems are gradually increasing. For example, some low earth orbit satellite communication systems and terrestrial microwave communication systems have gotten the transmission rate up to 10~2-10~3 Mbps. In such a high-speed wireless communication system, the wideband transmission channel's delay spread may cause intersymbol interference between hundreds of symbols. On the one hand, when the existing structure and algorithms of blind equalizers are used, a large number of taps are required to equalize the channel. On the other hand, the spectrum of the wideband channel will dramatically fluctuate in the passband. The two factors will cause slow convergence speed as well as high implementation complexity of the traditional blind equalizer, which doesn't meet the need of real-time processing applications. And so far, there is no mature idea or scheme to deal with these problems.In this dissertation, the structures and algorithms of wideband blind linear equalizer are deeply studied, to improve the convergence speed and real-time realization of blind equalization in high-speed, single carrier no-cooperative communication systems. The main contribution of this dissertation is that, based on the idea of subband adaptive filters, several novel wideband blind linear equalizers are proposed, which is suitable for single carrier communication system. In addition, the performance of these wideband linear blind equalizers is analyzed in detail.The main innovations of this dissertation are listed as follows. 1. Propose the uniform subband decomposition based wideband blind linear equalizerA new structure and algorithm of blind equalizer suitable for high-speed wideband communication are proposed by introducing uniform subband decomposition technique into the traditional blind linear equalizer. The proposed blind equalizer effectively combines the subband decomposition technique and sub-convolution method, and its convergence speed is faster than that of the traditional blind equalizer. On the other hand, parallel computation reduces the processing time, which is propitious to the algorithm's real-time realization.The tradition stop-and-go (SAG) algorithm is modified, according to property of the subband structure. It eliminated the memory error caused by subband decomposition, by separating the decision process into each subband. Experimental results prove that the convergence speed of the modified subband SAG algorithm is effectively improved.2. Propose a fractionally spaced wideband blind equalizerA fractionally spaced wideband blind equalizer is proposed by combining the fractionally spaced constant modulus algorithm (FSE-CMA) with the subband decomposition technique. The pre-whiten effect of the subband decomposition can significantly improve the convergence speed of FSE-CMA. Meanwhile, since the down-sampling and parallel processing of the subband decomposition as well as the sub-convolution are used, the proposed equalizer is beneficial to its real-time realization.3. Propose a non-uniform filter bank designed by the criterion of power spectrum amplitude segmentationIt is a novel method to design non-uniform filter bank based on the idea of power spectrum amplitude segmentation. The key feature is that it can effectively reduce the eigenvalue spread of auto correlation matrices of the subband signals, thus obviously improve the convergence speed of the subband least mean square (LMS) algorithm.The design method of the non-uniform filter banks is discussed in detail, including the algorithm of non-uniform subband allocation, the structures and the main design parameters of the filter bank. Theoretic analysis and simulation results show that the proposed non-uniform filter bank can effectively reduce the eigenvalue spread of auto correlation matrices of the subband signals, and possesses better performance of reconstructing the original signal4. Propose the non-uniform subband decomposition based wideband blind linear equalizerA structure and algorithm of wideband blind linear equalizer based on non-uniform subband decomposition is proposed, which utilizes the non-uniform subband decomposition method based on the idea of power spectrum amplitude segmentation. The main feature is that it can dynamically adjust the location and bandwidth of the subbands, and reduce the eigenvalue spread of auto correlation matrices of the subband signals more effectively than the one based on uniform subband decomposition. As a result, the convergence speed of the blind equalization algorithm can obviously be quickened.In addition, as the non-uniform subband signals have different sampling rates, an adapting strategy of the non-uniform subband equalizers is derived, which can effectively compromise between the convergence speed and computational complexity. By using the strategy, the equalizer can improve the convergence speed, and simultaneously reduce the computational complexity.