【作者】 韩钢；

【导师】 李建东；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2003， 博士

【摘要】 自适应传输技术是新一代移动通信和智能传输的核心技术之一。自适应调制是根据信道的实时状态以及业务的不同特性动态调整传输参数，从而可以充分挖掘系统的传输潜力，提高频谱利用率，以获得最大的传输容量和最高的可靠性。自适应传输中采用信号盲检测技术对信号的传输参数进行检测以实现收发双方的信息互通，可以节省信令的开销，对提高接收机的智能化水平有重要的研究意义。本文基于智能化传输的思想，对自适应调制中的信号盲检测技术进行了探索性的研究。本文的主要内容和成果如下： 1．研究了自适应调制中波特率，信噪比和载波相位等参数的估计算法；提出一种基于修改的欧几里得算法的波特率估计算法，这种算法可以对突发分组的波特率做出准确估计；提出一种星型QAM信号的信噪比估计算法，在中等信噪比条件下，具有较好估计性能。 2．对自适应单载波和多载波调制中的调制方式盲检测算法进行了研究；在自适应单载波调制中，研究了窄带信道下的AQAM调制原理和AQAM调制最佳星座图结构；在Rayleigh信道下和高斯信道下，提出一种基于高阶累积量的AQAM和ADPSK调制方式盲检测算法，算法具有良好的检测性能。对于自适应多载波调制，我们研究了OFDM调制和基于盲检测辅助的自适应OFDM的原理，提出一种在频率选择性衰落信道中AOFDM信号子信道调制方式盲检测算法，并与已有的文献进行了比较，证明了算法的有效性和稳健性。 3．研究了载波同步和码元定时同步与调制方式盲检测算法的关系；以自适应单载波中高阶累积量调制方式盲检测算法为例，对于载波同步误差引起的频偏问题，提出一种基于频偏稳健的MDPSK信号调制方式盲检测算法；对于未知调制方式信号的定时同步问题，提出一种盲定时估计算法，该算法可以估计MDPSK和MQAM信号的定时同步信息，实现数字信号的同步分类；提出了一种基于调制方式盲检测的自适应接收机结构，把调制方式盲检测，信噪比估计和同步解调联合起来进行，实现调制方式随信道质量而自适应变化的信号的正确接收。 4．在调制识别分类器的设计上，首次将统计学习理论的新成果——支撑矢量机应用在通信信号调制识别中；讨论了分类特征的选取，把接收信号的小波特征或高阶统计量特征作为识别特征，利用支撑矢量机分类器实现信号的调制识别。支撑矢量机把各个识别特征映射到一个高维空间，并在高维空间中构造最优识别超平面分类数据，实现通信信号的调制识别。该方法在信噪比变化范围较大的情况下，采用较少的训练数据就可以达到令人满意的识别正确率。更多还原

【Abstract】 As one of the key techniques of new generation mobile communication systems and broadband wireless communications systems, "Adaptive transmission" technique can exploit potential channel transmission ability sufficiently so that it enables the systems to reach the maximum transmission capacity and reliability by dynamically adjusting transmission parameters according to channel estimation and traffic QoS requirement.The most important character of the next generation mobile system will be smart transmission whose function module will be implemented by smart processing unit. This paper deals with the blind detection techniques in adaptive transmission. The main research works and results are listed as follows.1. Methods of estimating baud rate, Signal to Noise Ratio (SNR) and reference phase are investigated. A modified Euclidean algorithm is proposed to estimate baud rate of the burst packets. A cumulant based algorithm of estimating SNR of star-QAM is proposed which has better estimation performance in medium scope of SNR.2. The blind modulation detection algorithms in adaptive single- and multi-carrier modulation are studied. The principles of adaptive star-QAM and the optimum constellation of star QAM are given. In adaptive single-carrier modulation, a blind modulation detection algorithm based on Higher Order Cumulants (HOC) for AQAM and APSK is proposed. The excellent detection performance of algorithm is evaluated in Gaussian and narrow-band Rayleigh fading channel. In adaptive multi-carrier modulation, the principle of blind detection assisted adaptive OFDM (AOFDM) is given and a blind modulation detection algorithm is proposed for AOFDM in frequency selective fading channel. The performance of the algorithm is compared with the available algorithms and the effectiveness and robustness of the blind detection algorithm are proved through simulation.3. Take the HOC based blind modulation detection algorithm as an example, the relationship between the blind modulation detection algorithm and synchronization is investigated. An unproved detection algorithm robust to frequency offset is proposed which solves the problem caused by the error incarrier synchronization. How to synchronize a received signal with unknown modulation type is studied and a blind algorithm to estimate symbol timing of the signals with unknown modulation type is presented. The algorithm finds the best samples for the unknown received signals and realizes coherent recognition of modulations. An adaptive receiver structure is proposed which combines the blind modulation detection, SNR estimation, synchronization and demodulation. The receiver can properly receive the adaptive modulated signals.4. The design of classifier is dealt with and the support vector machine (SVM), a new result of statistics learning theory, is used in modulation recognition firstly. The choice of classification feature is analyzed. The classification feature vectors are extracted from Multi-level Wavelet Decomposition (MWD) and HOC of the received signals. SVM maps input vectors nonlinearly into a high dimensional feature space and constructs the optimum separating hyperplane in the spade to realize modulation recognition. Moreover, this method is robust to the variety of SNR and avoids overfitting and local minimum in neural nelwork. The percenlage of correcl idenlificalion for signals is salisfied wilh the fewer training data.更多还原