1-Bit压缩感知技术研究

发布时间：2018-05-23 09:25

本文选题：压缩感知 + 1-Bit压缩感知　；参考：《哈尔滨工业大学》2014年硕士论文

【摘要】：压缩感知（Compressive Sensing，CS）理论利用信号稀疏特性，克服了传统的奈奎斯特采样的限制，实现了信号的压缩、采样同时进行。该理论被广泛应用到模拟信息转换、光学成像、军事雷达及航空航天等领域。近几年压缩感知理论不断发展，并逐渐形成多种分支。1-Bit压缩感知理论作为压缩感知理论的分支，以其简单的结构和突出的重构效果受到研究人员的广泛关注，在无线通信和感知无线电等领域具有广泛的应用前景。 1-Bit压缩感知对采样值进行极限量化，并利用1-Bit重构算法恢复信号，，极大地简化硬件结构，降低采样值所占存储空间，提高信号存储和传输速率。但是该方向的研究起步较晚，且大多研究仅局限于理论层面，本文针对1-Bit压缩感知理论研究中重构算法存在信号稀疏度依赖问题和物理实现技术研究不充分的问题展开研究。主要研究内容和取得的成果如下： 1、研究1-Bit压缩感知基本理论。首先，介绍了1-Bit CS理论中1-Bit量化模型、观测矩阵约束条件和解的存在性与稀疏性等基本理论。接下来详细的介绍了1-Bit CS信号重构算法中较为突出的BIHT（Binary Iterative HardThresholding）算法，并给出了该算法的一系列衍生算法的重构模型和实施步骤，包括AOP（Adaptive Outlier Persuit）算法及其扩展算法。仿真实验表明在噪声的作用下，AOP算法和AOP-f（AOP-flips）算法的重构效果最好。 2、针对实际应用中信号稀疏度水平难以获取的问题，在现有的BIHT算法基础上进行修改，提出一种1-Bit压缩感知盲重构算法。首先对BIHT算法信号稀疏度依赖性问题进行研究；然后将稀疏度自适应思想引入BIHT算法中，提出1-Bit压缩感知盲重构算法；最后进行仿真实验，实验结果表明该算法在稀疏度未知的情况下，可以达到与BIHT算法相当的重构效果。 3、针对1-Bit压缩感知研究缺乏原型样机验证的问题，研究了一种基于1-Bit量化的多谐波信号采集系统。首先介绍多谐波信号模型。然后给出基于1-Bit量化多谐波信号采样系统框架，信号经过混频、滤波后进行1-Bit量化，得到采样值的符号信息。最后，完成系统硬件设计，分别给出各模块的设计方案，并进行硬件实验。实验结果说明基于1-Bit量化的多谐波信号采集系统可以成功恢复原信号，重构效果稳定，且在总Bits空间一定情况下，1-Bit量化的重构效果优于其它量化精度的重构效果。 4、针对稀疏多频带信号，提出一种基于1-Bit量化的多频带信号采集系统模型。首先描述了多频带信号模型，然后提出基于1-Bit量化的多频带信号采集系统框架，介绍了混频函数、频域变换和1-Bit量化模型，并提出适用于该系统的联合二进制迭代硬阈值-2范数算法，仿真实验结果表明在总Bits空间一定时，当输入噪声水平越高时，采用该系统进行1-Bit量化的重构效果较调制宽带转换器系统中采用精确量化的重构效果好，该系统表现出抗噪性、抗混叠、系统结构和实时性等工程方面的优势。
[Abstract]:Compressive Sensing (CS) theory uses signal sparsity to overcome the limitation of traditional Nyquist sampling and realizes the compression of signals and sampling simultaneously. This theory is widely applied to analog information conversion, optical imaging, military radar and aerospace. As a branch of compressed sensing theory, a variety of branch.1-Bit compression perception theory has been gradually formed, and its simple structure and outstanding reconstruction effect have been widely concerned by researchers. It has a wide application prospect in wireless communication and perceptual radio.
1-Bit compression perception quantifies the sampling value and uses the 1-Bit reconstruction algorithm to restore the signal, greatly simplifies the hardware structure, reduces the storage space occupied by the sampled values, and improves the storage and transmission rate of the signal. However, the research of this direction is late, and most of the research is limited to the theoretical level. This paper aims at the research of 1-Bit compression perception theory. The reconstruction algorithm has the problem of signal sparsity dependence and insufficient research on physical implementation technology. The main research content and achievements are as follows:
1, the basic theory of 1-Bit compression perception is studied. Firstly, the basic theories of 1-Bit quantization model in the 1-Bit CS theory, the existence and sparsity of the observation matrix constraints are introduced. Then, the more prominent BIHT (Binary Iterative HardThresholding) algorithm in the 1-Bit CS signal reconstruction algorithm is introduced in detail, and the algorithm is given. The reconfiguration model and implementation steps of a series of derivative algorithms, including the AOP (Adaptive Outlier Persuit) algorithm and its extended algorithm, are introduced. The simulation experiments show that the AOP algorithm and the AOP-f (AOP-flips) algorithm have the best reconfiguration effect under the effect of noise.
2, in view of the problem that the level of signal sparsity is difficult to obtain in practical application, a 1-Bit compression blind reconstruction algorithm is proposed based on the existing BIHT algorithm. First, the problem of signal sparsity dependence of BIHT algorithm is studied. Then, the sparse degree adaptive thinking is introduced into the BIHT algorithm, and the 1-Bit compression perception blindness is proposed. Finally, a simulation experiment is carried out. The experimental results show that the algorithm can achieve the same reconstruction effect as the BIHT algorithm when the sparsity is unknown.
3, in view of the lack of prototype prototype verification in 1-Bit compression perception research, a multi harmonic signal acquisition system based on 1-Bit quantization is studied. Firstly, the multi harmonic signal model is introduced. Then, the framework of the multi harmonic signal sampling system based on 1-Bit quantization is given. The signal is mixed frequency and 1-Bit quantization is carried out after the filter, and the symbol letter of the sampling value is obtained. Finally, the hardware design of the system is completed, the design scheme of each module is given and the hardware experiment is carried out. The experimental results show that the multi harmonic signal acquisition system based on 1-Bit quantization can successfully restore the original signal, and the reconfiguration effect is stable, and the reconstruction effect of the 1-Bit quantization is better than the other quantization precision in the case of the total Bits space. Construction effect.
4, a multi band signal acquisition system model based on 1-Bit quantization is proposed for sparse multiband signals. First, the multi band signal model is described. Then a multi band signal acquisition system framework based on 1-Bit quantization is proposed, and the mixing function, frequency domain transformation and 1-Bit quantization model are introduced, and the joint binary system is proposed. Iterative hard threshold -2 norm algorithm, the simulation results show that when the total Bits space is certain, when the input noise level is higher, the reconfiguration effect of 1-Bit quantization using this system is better than that in the modulation broadband converter system. The system shows the anti noise, anti aliasing, system structure and real-time performance. The advantage of the aspect.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7

【共引文献】