基于压缩感知的宽带信号采集与处理关键技术研究

发布时间：2018-07-18 09:24

【摘要】：随着无线通信技术的迅猛发展,电磁频谱环境日益复杂,在电磁频谱监测、无线电频谱感知、通信侦察等非合作接收应用中,不仅待处理的频带不断拓展,而且由于信号类型不断丰富,接收信号的动态范围也不断增大。因此,接收机必须具备大带宽、大动态的信号接收处理能力,而受到现有模拟数字转换(ADC)器件采样能力和动态范围水平的限制,以奈奎斯特理论为基础的宽带信号采集与处理机制面临着严峻的技术挑战。压缩感知技术针对稀疏信号或可压缩信号,以远低于奈奎斯特采样率的频率对信号进行压缩采样测量,降低了接收端对模拟数字转换器的要求,同时减轻了大数据量给后端存储和处理带来的压力,为解决宽带信号采集与处理中的难题带来了新的思路。本文围绕压缩感知理论应用于宽带信号采集与处理中的若干关键问题展开研究,重点针对感知矩阵不确定条件下的稀疏信号重构、基于压缩感知的宽带信号采集的动态范围分析、压缩域的干扰抑制、低信噪比下的宽带稀疏信号检测和调制识别以及基于压缩感知的电磁频谱监测方案实现等问题进行了研究,论文的主要工作和创新点如下:1、面向压缩感知在宽带信号采集中的实现,针对压缩感知中测量矩阵摄动和稀疏基矩阵失配的现实问题,建立感知矩阵不确定的压缩测量理论模型,提出一种基于误差限的联合正则化重构算法,理论分析和仿真结果表明,相对于标准的压缩测量模型和相应的重构算法,该算法在感知矩阵存在误差情况下,能够有效抵抗性能的恶化。标准压缩感知重构算法建立在理想的数学模型之上,但在测量矩阵摄动或稀疏基失配时,由于误差的影响将导致其重构性能恶化。本文在研究分析感知矩阵不确定因素的基础上,假定误差项有界,将感知矩阵不确定的信号重构问题转化为一个1和2范数联合约束的凸优化求解问题,通过1范数对重构信号的稀疏性进行约束的同时,引入2范数对感知矩阵不确定性进行约束,以一定的稀疏性为代价,实现稀疏信号有效重构的同时保证了求解的稳定性。2、针对宽带信号接收的无杂散动态范围性能问题展开研究,推导并给出正弦信号激励下ADC无杂散动态范围性能的理论界,分析和仿真该理论界和量化间隔、高斯噪声和采样率等因素的关系,得出相关结论;从理论上分析了压缩采样的量化噪声谱分布特征和无杂散动态范围性能,结果表明压缩采样的量化噪声谱分布和输入信号形式无关,相对于传统ADC采样,压缩采样的无杂散动态范围受噪声、ADC非线性等因素的影响较小,且可通过降低采样率、丢弃小部分饱和测量值等方式,有效提高宽带信号采集的动态范围。ADC无杂散动态范围测试方法受器件因素影响难以得到精确结果,本文在对ADC量化噪声谱分析基础上,理论推导了单音正弦信号激励下ADC的无杂散动态范围性能界,研究了量化比特、输入信号幅度、加性高斯噪声的方差等因素对SFDR性能的影响;结合傅里叶分析法,分析推导了不同采样率条件下的SFDR性能,得出了采样率和正弦激励信号频率呈“质数”关系时SFDR性能相对较好、整数倍采样时SFDR性能随采样率呈折线上升的结论;进一步对压缩测量的量化噪声谱进行了分析,得出了由于随机测量矩阵的作用,压缩感知的量化噪声谱是和输入信号形式无关的白噪声谱的结论;比较分析了ADC电路非线性对ADC采样和压缩采样的无杂散动态范围性能的影响,从降低采样率和测量值的公平性角度,阐明了压缩感知解决宽带信号采集中的大动态问题的优势。3、针对压缩感知框架下宽带信号的干扰抑制问题,提出了一种基于最小输出能量准则的压缩域干扰抑制算法,理论分析和仿真结果表明,该算法在干扰信号支撑集信息未知的情况下,能够有效抑制干扰对目标信号重构性能的影响。在压缩感知框架下,主要利用子空间正交投影算法和斜投影算法对干扰进行抑制,但都需要以干扰信号支撑集的先验知识为前提,这在非合作方的宽带信号采集与处理应用中通常无法满足。为此本文提出了一种无需支撑集先验知识的干扰抑制算法,该算法以感知矩阵每一列的期望投影的输出能量最小化为准则,设计相应的投影滤波器对压缩测量值进行投影滤波,进一步通过设定投影值门限对干扰信号进行抑制,同时保留目标信号的全部信息以便于后续的相关处理。4、针对宽带稀疏信号的检测和调制识别问题,将循环谱的相关理论引入压缩感知框架下,基于循环频率切面的近似块稀疏特性,提出一种基于压缩域循环谱能量特征的信号检测算法,仿真结果表明,该算法能够有效实现较低信噪比条件下的稀疏信号检测。以此为基础设计了一种基于二分迭代的压缩域循环谱特征提取方法,并结合二叉树分类器,实现了调制信号的识别。首先分析了现有压缩感知框架下,经典的压缩检测、子空间检测算法在低信噪比条件下检测性能的局限性,然后基于大部分调制信号具备循环平稳特性以及高斯白噪声只在零循环频率处出现的事实,将循环谱分析引入到压缩感知框架下,提出基于压缩域循环谱能量特征的稀疏信号检测算法。该算法不同于现有压缩循环谱的信号检测算法,不需要对压缩循环谱进行完全重构,且充分利用信号在循环频率切面的近似块稀疏特性,所需的压缩测量数目和计算量大大降低,仿真结果表明该算法在低信噪比条件下能有效实现信号的检测。最后在块稀疏压缩循环谱模型的基础上,给出一种基于二分迭代的循环谱特征提取方法,并结合二叉树分类器实现对{BPSK,FSK,2ASK,16QAM,MSK}五类常见信号的调制识别。5、针对“电磁频谱监测传感器网络关键技术研究”课题中电磁频谱监测的需求,设计了基于AIC压缩测量的电磁频谱监测方案和原理验证平台,并针对AIC实现时滤波器非理想导致重构性能恶化的问题,提出了一种基于自适应滤波校正的方法,提高了重构性能。为降低瞬时突发信号的漏检概率,缓解前端ADC的压力和要求,本方案基于AIC的压缩测量思想对电磁频谱信号进行宽带采集,并利用测量值直接在压缩域实现信号检测、调制识别等信号处理工作。设计实现了方案的原理验证平台,模拟端采用RD结构,数字端考虑到系统的可扩展性和灵活性,采用GPU+FPGA+ARM的结构。最后针对基于RD的模拟压缩测量实现过程中,滤波器冲激响应非理想影响信号重构性能的问题,设计自适应滤波校正算法对非理想滤波器的脉冲冲激响应进行估计,提高了系统的性能,且无需改变原有压缩测量结构。
[Abstract]:With the rapid development of wireless communication technology, the electromagnetic spectrum environment is increasingly complex. In the non cooperative receiving applications, such as electromagnetic spectrum monitoring, radio spectrum sensing, communication reconnaissance and other non cooperative applications, the frequency band of the processing is not only expanded, but also the dynamic range of received signals is increasing because of the continuous enrichment of signal types. Therefore, the receiver must be equipped with With large bandwidth, large dynamic signal reception and processing capability, and limited by the existing analog digital conversion (ADC) device sampling and dynamic range level, the wideband signal acquisition and processing mechanism based on Nyquist theory faces severe technical challenges. Compression sensing technology is far lower for sparse signal or compressible signal. The compression sampling measurement of the Nyquist sampling rate reduces the demand for analog digital converter at the receiving end, reduces the pressure caused by the large amount of data to the back end storage and processing, and brings new ideas to solve the difficult problems in the acquisition and processing of wide-band signals. Several key problems in signal acquisition and processing are studied, focusing on sparse signal reconstruction under the uncertainty of perceptual matrix, dynamic range analysis of compressed sensing based wideband signal acquisition, interference suppression in compressed domain, wide-band sparse signal detection and modulation recognition under low signal to noise ratio, and electricity based on compressed sensing. The main work and innovation of this paper are as follows: 1, for the realization of compressed sensing in the acquisition of wide-band signal, in view of the real problems of the measurement matrix perturbation and the sparse matrix mismatch in the compressed sensing, a theoretical model of the uncertainty of the perception matrix is set up, and a kind of theory based on the uncertainty of the perception matrix is proposed. The theoretical analysis and simulation results show that, compared with the standard compression measurement model and the corresponding reconstruction algorithm, the algorithm can effectively resist the deterioration of the performance when the perceptual matrix exists error. The standard compression perception reconstruction algorithm is based on the ideal mathematical model, but the measurement moment is in the measurement moment. In this paper, on the basis of analyzing the uncertainty factors of the perception matrix, this paper assumes that the error term is bounded, and transforms the signal reconstruction problem of the uncertainty of the perception matrix into a convex optimization problem with a combination of 1 and 2 norm constraints, and the reconstruction of the reconstructed letter through the 1 norm. When the sparsity of the number is constrained, the 2 norm is introduced to restrict the uncertainty of the perceptual matrix. At the expense of a certain sparsity, the effective reconstruction of the sparse signal is realized and the stability of the solution is guaranteed at the same time. The study on the performance of the non stray dynamic range of the wideband signal receiving is studied, and the sinusoidal signal excitation is derived and given under the excitation of the wideband signal receiving. ADC has no theoretical circle of stray dynamic range performance, analyzes and emulates the relationship between the theoretical circle and the quantization interval, the relationship between the Gauss noise and the sampling rate, and draws the relevant conclusions. The quantitative noise spectrum distribution and the non stray dynamic range performance of the compressed sampling are analyzed theoretically. The results show the quantization noise spectrum distribution and input of the compressed sampling. The signal form is independent. Compared with the traditional ADC sampling, the non stray dynamic range of the compressed sampling is less affected by the noise, the ADC nonlinearity and other factors, and it can effectively improve the dynamic range of the wideband signal acquisition by reducing the sampling rate and discarding the small part of the saturation measurement. The method is affected by the device factors in the dynamic range of the wideband signal acquisition.ADC. It is difficult to obtain accurate results. On the basis of ADC quantization noise spectrum analysis, this paper derives the performance boundary of ADC without stray dynamic range under the excitation of monosyllabic sinusoidal signal, and studies the influence of quantized bits, input signal amplitude, variance of additive Gauss noise and other factors on the performance of SFDR. The SFDR performance under the sample rate is obtained. It is concluded that the SFDR performance is relatively good when the sampling rate and the sinusoidal excitation signal frequency is "prime", and the SFDR performance increases with the sampling rate when the integer multiple sampling is sampled. The quantization noise spectrum of the compression measurement is further analyzed, and the compression perception is obtained due to the effect of the random measurement matrix. The quantization noise spectrum is the conclusion of white noise spectrum unrelated to the input signal form. The influence of ADC circuit nonlinearity on the non stray dynamic range performance of ADC sampling and compressed sampling is compared and analyzed. From the angle of reducing the sampling rate and the fairness of the measured values, the advantage.3 of the compressed sensing to solve the big dynamic problems in the broadband signal acquisition is clarified. In view of the interference suppression of wide-band signals under the compressed sensing framework, a compression domain interference suppression algorithm based on the minimum output energy criterion is proposed. The theoretical analysis and simulation results show that the algorithm can effectively suppress the influence of interference on the performance of target signal reconstruction under the condition of unknown interference signal support set. Under the framework of contraction sensing, the interference is suppressed mainly by subspace orthogonal projection algorithm and oblique projection algorithm, but all of them need to be based on the prior knowledge of the interference signal support set, which is usually not satisfied in the application of non cooperative wideband signal acquisition and processing. The algorithm is based on the minimization of the output energy of the expected projection of each column of the perceptual matrix, and the corresponding projection filter is designed for the projection filtering of the measured value, and the interference signal is suppressed by setting the threshold value of the projection value, and all the information of the target signal is retained to facilitate the subsequent related processing of.4. In view of the detection and modulation recognition of wide-band sparse signals, the correlation theory of cyclic spectrum is introduced into the compressed sensing framework. A signal detection algorithm based on the cyclic spectral energy characteristics of the compressed domain is proposed based on the sparse characteristic of the approximate block in the circular frequency section. The simulation results show that the algorithm can effectively implement the low signal to noise ratio conditions. On the basis of this, we design a cyclic spectral feature extraction method based on two sub iteration, and combine the two forked tree classifier to realize the modulation signal recognition. First, it analyzes the local detection performance under the existing compression sensing framework, the classic compression detection and subspace detection under the low signal to noise ratio conditions. Based on the fact that most modulation signals have cyclostationary characteristics and the fact that Gauss white noise appears only at zero cycle frequency, the cyclic spectrum analysis is introduced into the compressed sensing framework, and the sparse signal detection algorithm based on the cyclic spectral energy characteristics of the compressed domain is proposed. This algorithm is different from the existing compression cycle spectrum signal detection. The algorithm does not need to reconstruct the compression cycle spectrum completely, and make full use of the approximate block sparsity of the signal in the circular frequency section. The number of compressed measurements and the amount of computation are greatly reduced. The simulation results show that the algorithm can detect the signal effectively under the condition of low signal to noise ratio. Finally, the base of the block sparse compression cyclic spectrum model is based on the simulation results. On the base of this, a cyclic spectral feature extraction method based on two minute iteration is given, and the modulation recognition.5 for the common signals of {BPSK, FSK, 2ASK, 16QAM and MSK} is realized with the two fork tree classifier. According to the requirement of the electromagnetic spectrum monitoring in the key technology research of the electromagnetic spectrum monitoring sensor network, the AIC compression measurement is designed. The electromagnetic spectrum monitoring scheme and the principle verification platform, and aiming at the problem that the non ideal filter results in the deterioration of the reconstruction performance when the AIC is realized, a method based on adaptive filtering correction is proposed to improve the reconstruction performance. In order to reduce the leakage probability of the instantaneous burst signal and alleviate the pressure and requirement of the ADC in the front end, the scheme is based on the compression of AIC. The idea of measuring the wide-band signal of the electromagnetic spectrum signal, and using the measured value directly in the compressed domain to realize signal detection, modulation recognition and other signal processing work. The design and Realization of the scheme's principle verification platform, the analog end uses the RD structure, the digital end takes into account the extensibility and flexibility of the system, and uses the structure of the GPU+FPGA+ARM. Finally the needle is used. In the implementation of RD based analog compression measurement, the impulse response of the filter is not ideal for the performance of the signal reconstruction. The adaptive filter correction algorithm is designed to estimate the impulse impulse response of the non ideal filter, which improves the performance of the system and does not need to change the original compression measurement structure.
【学位授予单位】：解放军信息工程大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN911.7

【参考文献】