【作者】 靳晓艳；

【导师】 周希元；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2014， 博士

【摘要】 通信信号调制识别是指对接收信号自动处理并判定其调制类型的过程。作为信号检测与解调的中间环节，调制识别技术在认知无线电、智能解调器、电子侦察等各种民用及军事应用中扮演着重要角色。调制识别技术经过几十年的发展，虽然已经取得了很多成果，但随着工程化需求逐渐提高，无线通信信道环境日益复杂，仍有不少问题亟待解决。本文重点致力于似然比调制识别算法的工程化应用和基于先进信号处理方法的调制识别技术的研究，论文的主要工作和创新性成果主要包括：1.从通信信号处理的工程实用角度出发，针对似然比调制识别算法计算复杂度高的问题，提出了一种改进的快速算法。该算法引入硬件实现中预存查表的思想，通过一个查找表的地址读出待识别信号的似然函数值以便节省在线处理时间，为似然比调制识别算法的实时化应用提供了一种解决思路。试验结果表明算法能够在保证似然比算法最优性的基础上有效地节约时间成本，更能适应实时性要求高的应用场合。2.在平坦衰落信道环境下，提出了基于一种自适应马尔可夫链蒙特卡罗（MCMC）技术——自适应Metropolis（AM）技术的调制识别算法。该算法能够在迭代过程中产生满足目标分布的未知参数和发送符号的各态历经样本，从而在实现似然函数的近似计算的同时完成参数估计，可实现调制识别与参数估计的一体化处理。相比于传统Metropolis-Hastings（MH）技术，AM技术避免了因建议分布函数选取不当造成的性能损失。仿真结果表明在平坦衰落信道环境下，基于AM技术的调制识别算法能够快速、精确的收敛，具有很好的识别性能。3.将平坦衰落信道环境中基于AM技术的调制识别算法推广到多径信道环境中，为解决AM算法在未知参数维数较高时收敛速度放缓的问题，提出了基于单分量自适应Metropolis（SCAM）技术的调制识别算法。该算法在迭代过程中按顺序依次对未知参数向量中每个分量单独进行类似AM采样的操作。仿真结果表明在未知参数维数较高的情况下，SCAM算法拥有更加优越的收敛性能，基于SCAM技术的调制识别算法在多径信道环境下识别性能良好。4.在低信噪比环境下，提出了基于混沌理论的调制识别算法。文中对Duffing振子大尺度周期态特性进行了系统研究，考察了激励信号频率、幅度和相位对Duffing振子周期解的影响。根据大尺度周期状态下系统解随激励信号相位变化的规律，设计了基于Duffing振子的MPSK信号调制识别算法。并根据调制识别中的特征提取任务需求，对Duffing振子进行了模型优化，将优化后Duffing振子系统解的Poincaré映射作为分类特征，实现了低信噪比下的MPSK信号的调制识别。该算法仅仅利用载波频率的先验信息，无需进行码元同步，并且对信号幅度、载波初始相位以及载波频率的变化不敏感。仿真表明该算法具有较强的噪声免疫力，在信噪比较低的情况下仍能达到满意的识别效果。更多还原

【Abstract】 Modulation recognition of communication signal is referred to the process ofhandling the received signal and judging its modulation type automatically. As theintermediate link between signal detection and demodulation, modulation recognitiontechnique plays an important role in cognitive radio, smart modem, electronicsurveillance and other civilian and military applications. After decades of development,modulation recognition technique has got many achievements, but with the demand forengineering increasing gradually, and wireless communication channel environmentbecoming more complex, there are still many problems needed to be solved formodulation recognition.This paper focuses on the engineering applications for likelihood ratio modulationrecognition algorithm and the research of modulation recognition technique based onadvanced signal processing methods. The main work and innovative achievementsobtained in this paper are summarized as follows:1. For the purpose of engineering practicality for communication signalprocessing, an improved fast algorithm is proposed to solve the problem of highcomputational complexity for the likelihood ratio modulation recognition algorithm.The idea of pre-stored and look-up table in hardware implementation is introduced tothe algorithm. In order to save online processing time, read the likelihood function valueof the signal to be identified through a table address, which providing a solution idea forthe likelihood ratio modulation recognition algorithm for real-time applications.Experimental results show that the fast algorithm can guarantee optimality of thelikelihood ratio based algorithm, at the same time it can effectively save time cost andadapt to the applications with high real-time demands.2. A modulation recognition algorithm based on a kind of adaptive Markov chainMonte Carlo (MCMC)——adaptive Metropolis (AM) technique is proposed in flatfading channel environment. The algorithm generates ergodic samples of the unknownparameters and transmitted symbols in the iterative process to meet the targetdistribution, which can realize approximate calculation of the likelihood function andcomplete parameter estimation meanwhile. So the algorithm can achieve the integrationof modulation recognition and parameter estimation. Compared to the traditionalMetropolis-Hastings (MH) technique, AM technique can avoid the performance losscaused by improper selection of the proposed distribution function. Simulation results show that in flat fading channel environment, the proposed modulation recognitionalgorithm based on AM technique can converge quickly and accurately, and it has goodrecognition performance.3. The modulation recognition algorithm based on AM technique in flat fadingchannel environment is extended to the multi-path channel environment. Theconvergence rate of AM algorithm will slow down when the unknown parametersvector has high dimension, to solve this problem, a modulation recognition algorithmbased on single-component adaptive Metropolis (SCAM) technique is proposed. Thealgorithm operates AM sampling on each single component in unknown parametervector successively in the iterative process. Simulation results show that in the case ofhigh dimension unknown parameters, SCAM algorithm has superior convergenceperformance, SCAM technique based modulation recognition algorithm in multi-pathchannel environments can classify modulated signals effectively.4. A modulation recognition algorithm based on chaos theory is proposed in lowSNR environment. The properties of Duffing oscillator in large-scale periodic state arestudied systematically in this paper, the effections of frequency, amplitude and phase ofthe excitation signal on the periodic solutions for Duffing oscillator are investigated.According to the rule of system solution in large-scale periodic state changing with thephase of the excitation signal, a modulation recognition algorithm is designed based onDuffing oscillator for MPSK signals. The Duffing oscillator model is optimized basedon requirements for feature extraction in modulation recognition, the Poincaré map ofoptimized Duffing oscillator model acts as classification features to achieve MPSKsignal modulation recognition in low SNR. The algorithm only uses the prioriinformation of carrier frequency, it can be implemented without symbol synchronization,and it is insensitive to variations of the signal amplitude, the carrier initial phase and thecarrier frequency. Simulation results show that the algorithm has strong noise immunity,it can achieve satisfactory recognition results in the case of low SNR environment.更多还原