Research On Modulation Classification, Channel Blind Identification And Equalization In Wireless Fading Channels

The complexity of wireless channels and the variety of radio communications give more challenge to blind receiver. Especially, the received signal is distorted by channel fading, which causes more difficulties for correct demodulation in uncooperative receiver. The automatic modulation classification of communication signals and the blind identification and blind equalization of wireless channels are key technologies for effective interception in fading channels. The former is the precondition of correct demodulation, and the latter is an effective solution to overcome the influence of channels. Both of them are independent and interdependent. In this dissertation, we study the two key issues in fading channels.Firstly, the modulation classification between orthogonal frequency division multiplexing (OFDM) signal and digital single-carrier modulated signals in fading channels is studied. To the best of our knowledge, there has no research on the issue in published files. The modulation classification algorithm in Rayleigh fading channels is proposed, in which the modulation classification feature is deduced and the joint signal-to-noise ratio estimation and modulation classification method is used as classifier. Moreover, the idea is extended to the classification between OFDM and single-carrier modulation types in "bad" channels of short wave communication. A large number of simulations show that, with enough received samples (>2000) and enough sub-carriers of OFDM (>4), the correct classification probability is high.Secondly, the classification of higher order QAM signals in linear time-invariant fading channels is addressed in detail and the joint equalization and modulation classification algorithm is proposed. On the basis of the existing double-mode equalization methods, an improve HY-NCMA algorithm is proposed as the core to construct the equalizer bank. To implement the joint equalization and classification, the mean square error of the output from each equalizer is calculated as classification feature, and the minimum mean square error criteria is used as classifier. Unlike the conventional two-step algorithm, the new algorithm takes the full advantage of the equalization and avoids the complex classification stage. So, not only are the sample size and the complexity of the proposed algorithm reduced, but also the identification rate is improved. Simulation results demonstrate the efficiency of the modulation classification.Thirdly, the dissertation studies the classification of MPSK and MQAM signals based on the blind channel identification and blind equalization using the second-order statistics when the linear time invariant channels experience a deep fading. A subspace based modulation classification algorithm is proposed with knowing channel order. Compared with the existing methods, the subspace based algorithm can classify more modulation types and improve the classification performance. In fact, the accurate channel order is not known previously. From the point of view, the channel order estimation algorithms are studied and the improved channel order estimation methods are proposed. Furthermore, the improved outer-product decomposition algorithm (OPDA) algorithm with better robustness is proposed, which needs the knowledge of upper bound of the channel order, but the exact channel order. On the same circumstance, the modulation classification of MPSK and MQAM signals based on the improved OPDA is proposed by combining the improved OPDA algorithm with the multi-step sub-cluster method, to the best of our knowledge, which is the first time to implement modulation classification without prior knowledge of exact channel order. Then, the improved channel order estimation method is used in the subspace based modulation classification algorithm for the same end, whereafter which is compared with the improved OPDA based algorithm. The performances of those algorithms are proved by a lot of simulations.Finally, as the key technology of modulation classification in linear time varying (LTV) channels, the blind channel identification of the LTV channels is studied. On the basis of the discrete-time canonical model, the two-stage blind identification algorithm based on the improved OPDA is proposed for the double-selective channels. Compared with existing methods, the new algorithm is more practical because it does not need the prior knowledge of the real channel order but the upper bound.