【作者】 刘明骞；

【导师】 李兵兵；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2013， 博士

【摘要】 信号检测及调制识别技术是非合作通信中接收机的关键技术,在自适应调制、军事侦察、电子对抗中有重要的应用价值。针对目前非合作通信中信号的检测与识别性能低且不能满足实际应用需求的问题,本文对非合作通信中信号检测及调制识别的关键技术进行了系统的研究,主要涉及到多径信道下数字调制信号的参数估计和识别、非高斯噪声下数字调制信号的识别以及单通道时频重叠信号的参数估计和识别,所取得的主要研究成果为：1.针对传统的OFDM信号带宽估计方法在低信噪比多径信道下估计精度低且计算量大的问题,提出一种基于经验模态分解的OFDM信号带宽估计方法,该方法有效地克服了噪声的影响。仿真结果表明,该方法不但估计精度更高且计算复杂度更低；针对多径信道下传统的OFDM信号信噪比盲估计方法的估计性能差、计算复杂度高的问题,提出了一种基于AR模型和高阶累积量的OFDM信号信噪比盲估计方法,该方法无需任何先验信息,在多径信道下具有良好的估计性能且计算复杂度低。针对低信噪比多径信道下ZP-OFDM信号时域参数估计精度低的问题,提出了一种改进的ZP-OFDM信号的时域参数盲估计方法,该方法通过功率自相关函数和小波消噪及平滑滤波对符号总长度进行了有效地估计,并利用符号总长度数据的幅度平方值和小波分解与重构对零前缀估计长度也进行了估计。仿真结果表明,在低信噪比多径信道条件下,该方法具有良好的估计性能。2.针对低信噪比多径信道下含有频偏和相偏的数字调制信号识别性能差的问题,提出了一种基于循环平稳特性的数字调制信号识别方法,不仅从理论上证明了该方法可以消除频偏、相偏、噪声和多径衰落信道的影响,并通过仿真验证了该方法的有效可行性和稳健性；针对多径信道下含有导频的OFDM信号子载波调制方式难识别的问题,提出了一种基于高阶累积量和改进的减法聚类相结合的OFDM信号子载波识别方法,该方法可以消除多径信道带来的衰减和相位旋转的影响。结果表明,在多径环境下该方法有效地实现了含有导频的OFDM信号子载波调制方式的识别。3.针对传统数字调制识别方法在非高斯噪声下识别性能差的问题,该文提出一种基于广义分数阶傅里叶变换和分数低阶Wigner-Ville分布的数字调制识别方法,该方法提取广义分数阶傅里叶变换的零中心归一化瞬时幅度谱密度的最大值和分数低阶Wigner-Ville分布幅度的最大值作为识别特征参数有效地实现了非高斯噪声下数字调制信号识别。仿真结果表明,在非高斯噪声下,该识别方法不仅性能明显优于传统方法并且具有较高的识别率和良好的稳健性。针对传统数字调制识别方法在Alpha稳定分布噪声下识别性能差、计算复杂度高的问题,本文提出了一种基于广义分数阶傅里叶变换和分数低阶协方差谱的的数字调制识别方法。该方法提取广义分数阶傅里叶变换的归一化瞬时频率谱密度的最大值和分数低阶协方差谱幅度的最大值作为识别特征参数,有效地实现了非高斯噪声下数字调制信号识别。仿真结果表明,在Alpha稳定分布噪声下,该方法不但具有较高的识别率和良好的稳健性并且计算复杂度更低。4.针对传统的单通道时频重叠信号的个数估计方法性能差的问题,提出了一种基于四阶循环累积量的单通道时频重叠信号的个数估计方法,该方法不但具有良好的抗噪性能,并且对信号频谱重叠率具有鲁棒性；针对传统的单通道时频重叠信号的载波频率估计方法性能差、计算复杂度高的问题,提出了一种基于四阶循环累积量的单通道时频重叠信号的载波频率估计方法。仿真结果表明,该方法在低信噪比条件下具有良好的估计性能且计算复杂度低；针对传统的单通道时频重叠信号的码元速率估计方法性能差、计算复杂度高的问题,提出了一种基于循环谱的单通道时频重叠信号的码元速率估计方法,该方法从理论上分析了在循环谱截面上进行离散谱线检测的可行性,并通过计算机仿真验证了其有效性。仿真结果表明,该方法不但具有较强的抗噪能力并且具有更低的计算复杂度。5.针对单信道时频重叠信号的识别存在识别性能差、识别调制类型单一的缺点,提出了一种基于高阶循环累积量的单通道时频重叠信号识别方法,该方法具有较强的抗噪能力。仿真结果表明,该方法在低信噪比条件下不但识别性能好、计算复杂度低,而且对信号频谱重叠率具有鲁棒性。更多还原

【Abstract】 Signals detection and modulation recognition are the key technologies of receivers in non-cooperative communication system, which have the important application value in adaptive modulation, military reconnaissance and electronic countermeasure fields. According to the problems that current signals detection and modulation recognition technologies have poor performance and could not meet the requirement of practical application in non-cooperative communication. This dissertation is mainly concerned with the the key technologies of signals detection and modulation recognition in non-cooperative communication, which including parameters estimation and recognition of digital modulation signals over multipath channels, recognition digital modulation signals with non-Gaussian noise and parameters estimation and recognition for time-frequency overlapped signals over single channel. The author’s major contributions are outlined as follows:1. According to the problems that traditional bandwidth estimation schemes for OFDM signals have low estimation accuracy and high complexity in low SNR over multiple channels, a novel bandwidth estimation scheme for OFDM signals base on Empirical Mode Decomposition is proposed. This scheme effectively overcomes the effects of noise. Simulation results show that this scheme has higher estimation accuracy and lower computational complexity. According to the problems that traditional SNR blind estimation schemes for OFDM signals have poor performance and high1complexity over multiple channels, a novel SNR blind estimation scheme for OFDM signals based on AR model and high-order cumulant is proposed and this scheme dose not need any priori information and has better estimation performance and lower complexity over multiple channels. According to the problems that time domain parameter estimation schemes for ZP-OFDM signals have low estimation accuracy in low SNR over multiple channels, a novel time domain parameter estimation scheme for ZP-OFDM signals is proposed. This scheme effectively estimates the total length of the symbols by power autocorrelation function, Wavelet de-noising and smoothing filtering. This scheme also estimates the length of the zero prefix by using magnitude-squared value of the data of the total length of the symbols and Wavelet decomposition and reconstruction. Simulation results show that this scheme has better performance in low SNR over multiple channels.2. According to the problems that the recognition schemes of digital modulation signals with frequency offset and phase offset have poor performance in low SNR over multipath channels, a novel recognition scheme of digital modulation signals base on cyclostationarity property is proposed. It is proved that this scheme could eliminate the effect of frequency offset, phase offset, noise and multipath fading channels. Simulation results show that this scheme has effective feasibility and robustness. According to the problems that the modulation subcarriers of OFDM signals with pilot are recognized difficultly over multiple channels, a novel subcarriers recognition scheme for OFDM signals base on high-order cumulant and improved subtractive clustering is proposed. This scheme could eliminate the effect of attenuation and phase rotation. Simulation results show that this scheme could effectively achieve the recognition of modulation subcarriers for OFDM signals with pilot.3. According to the problems that traditional recognition schemes of digital modulation signals with non-Gaussian noise have poor recognition performance, this paper is proposed a novel recognition scheme based on generalized Fractional Fourier Transform and fractional lower Wigner-Ville distribution. This scheme extracts the recognition characteristic parameters which are the maximum of the spectrum density of zero center normalized instantaneous amplitude of Fractional Fourier Transform and the maximum of fractional lower Wigner-Ville distribution amplitudes to identify the digital modulation signals with a non-Gaussian noise. Simulation results show that this scheme has better recognition performance and better robustness than the traditional schemes. According to the problems that traditional recognition schemes of digital modulation signals have poor recognition performance and high complexity in an Alpha stable distribution noise environment, a novel recognition scheme of digital modulation signals based on generalized Fractional Fourier Transform and fractional lower covariance spectrum is proposed. This scheme extracts the characteristic parameters which are the maximum of the spectrum density of normalized instantaneous frequency of Fractional Fourier Transform and the maximum of fractional lower covariance spectrum amplitudes to identify the digital modulation signals an Alpha stable distribution noise environment. Simulation results show that this scheme has higher recognition rates, better robustness and lower complexity than the traditional schemes.4. According to the problems that traditional number estimation schemes for time-frequency overlapped signals over single channel have poor performance, a novel number estimation scheme for overlapped signals based on fourth-order cyclic cumulant is proposed. This scheme has good anti-noise performance and is robust for signals spectrum overlap rates. According to the problems that traditional frequency estimation schemes for time-frequency overlapped signals over single channel have poor performance and high complexity, a novel frequency estimation scheme for overlapped signals based on fourth-order cyclic cumulant is proposed. Simulation results show that this scheme has better estimation performance and lower complexity than the traditional schemes. According to the problems that traditional symbol rate estimation schemes for time-frequency overlapped signals over single channel have poor performance and high complexity, a novel symbol rate estimation scheme for overlapped signals based on cyclic spectrum is proposed. This scheme analyzes the feasibility of discrete spectral lines detection in the cyclic spectrum cross-section and verifies its effectiveness in simulation. Simulation results show that this scheme not only has strong anti-noise ability and lower complexity.5. According to the problems that the traditional recognition schemes for time-frequency overlapped signals over single channel have poor performance and only can recognize single type modulation signal, a novel recognition scheme for time-frequency overlapped signals over single channel based on high-order cyclic cumulant is proposed. This scheme has strong anti-noise ability. Simulation results show that this scheme not only has better recognition performance and lower complexity, but also is robust for signals spectrum overlapped rates.更多还原