压缩传感理论应用于无线传感器网络关键技术研究

发布时间：2018-06-11 17:04

本文选题：压缩传感 + 无线传感器网络　；参考：《南开大学》2014年博士论文

【摘要】：作为物联网的核心技术,无线传感器网络(WSN)因为其应用灵活性和信息感知有效性受到了越来越多的关注。也正是因为要保证传感器节点的灵活性,其硬件资源和能源供给部分受到限制,难以满足大规模高密度海量信息的传送和处理,成为制约WSN大规模应用的重大技术难题。近年来,压缩传感(CS)理论获得了广泛关注和研究。压缩传感将采样与压缩过程合并,直接将稀疏或者可压缩信号中的“冗余”信息丢弃,因此降低了信号采样频率,并且节省了存储和传输成本。压缩传感理论的出现,为无线传感器网络的海量数据采集、传输、存储以及节点续航能力等问题提供了一种全新的技术解决方案,可加快物联网工程发展的步伐。本文在对无线传感器网络数据特性分析的基础上,将压缩传感理论的关键技术作为主要研究内容,致力于将CS理论应用到WSN中,为此展开了多方面研究工作： 1.信号稀疏表示：在自然界中的稀疏信号是少见的,但是大部分信号都可以在某个域上进行稀疏表示。在深入研究WSN数据特性的基础上,利用过完备字典对信号进行稀疏,分别生成了离散余弦基、Haai、Chirplet以及Db小波等过完备原子库,并进行了多参量级联过完备学习字典仿真,仿真结果表明：Chirplet和Db小波过完备字典稀疏效果优于DCT过完备字典和Haar小波过完备字典,使用范围也更广；级联过完备字典对于与先验模型相似的信号稀疏效果非常明显。 2.测量矩阵：基于满足约束等距性(RIP)这一条件,对高斯随机测量矩阵、伯努利随机测量矩阵和托普利兹以及循环测量矩阵等进行了研究,并通过仿真对之进行了对比分析。并在此基础上,提出了一种易于硬件实现、存储空间需求低的伯努利伪随机循环矩阵。仿真结果表明：在测量数M满足一定条件时,伯努利伪随机矩阵可以高精度实现信号的测量与重构。 3.信号重构：重构算法是目前研究较深入而且成果较多的一项技术,本文从重构精度、速度以及成功率等方面对现有的重构算法进行了对比分析,在分析现有各种算法优劣性的基础之上,根据WSN的数据特性,提出了一种实用性更强、重构精度更高、稳定性和鲁棒性更好的ITSAOMP重构算法。实验结果表明：在测量矩阵满足一定条件时,借助于ITSAOMP算法,可以高概率、低失真地重构原始信号,并且具备较好的噪声鲁棒性。 4.将压缩传感和周期非均匀采样有机结合：周期非均匀采样是有效降低采样频率、提高采样精度的一种方法,它利用多通道采样系统对信号进行采样。根据非均匀采样系统的特点,利用联合子空间理论将采样和重构过程转化为矩阵或向量运算,并借助CS理论,将稀疏信号重构算法应用到非均匀采样系统中对信号进行重构,仿真结果表明系统不仅很好地实现信号采样与重构,而且大大降低了采样频率,提高了重构精度。
[Abstract]:As the core technology of the Internet of things, Wireless Sensor Networks (WSNs) has attracted more and more attention for its flexibility in application and effectiveness of information perception. It is also because the flexibility of sensor nodes is guaranteed that the hardware resources and energy supply are limited, so it is difficult to meet the transmission and processing of large-scale high-density and magnanimous information. In recent years, the theory of compressed sensing (CSN) has received extensive attention and research. The compression sensor combines the sampling process with the compression process and directly discards the "redundant" information in the sparse or compressible signal, thus reducing the sampling frequency and saving the storage and transmission costs. The emergence of compressed sensing theory provides a new technical solution for mass data acquisition, transmission, storage and node endurance of wireless sensor networks. Based on the analysis of the data characteristics of wireless sensor networks, the key technology of compression sensing theory is taken as the main research content in this paper, and the CS theory is applied to WSN. To this end, a number of research work has been carried out: 1. Signal sparse representation: sparse signals are rare in nature, but most signals can be represented sparsely in a certain domain. On the basis of deeply studying the characteristics of WSN data, the over complete dictionary is used to sparse the signal, and the over complete atomic libraries such as the discrete cosine base Haaian Chirplet and the Db wavelet are generated, respectively, and the simulation of the multi parameter cascade over complete learning dictionary is carried out. The simulation results show that the sparse effect of the over complete dictionaries of the two wavelets is better than that of the over complete dictionaries of DCT and Haar wavelets, and the cascaded over complete dictionaries have obvious effect on the sparsity of signals similar to the prior models. 2. Measurement matrix: based on the condition of satisfying the constraint equidistant property, the Gao Si random measurement matrix, Bernoulli random measurement matrix, Topril matrix and cyclic measurement matrix are studied, and the simulation results are compared and analyzed. On this basis, a Bernoulli pseudorandom cyclic matrix is proposed, which is easy to implement in hardware and low in storage space. The simulation results show that the Bernoulli pseudorandom matrix can be used to measure and reconstruct the signal with high precision when the measurement number M satisfies certain conditions. Signal reconstruction: the reconstruction algorithm is a technology which has been studied deeply and has achieved a lot at present. In this paper, the existing reconstruction algorithms are compared and analyzed from the aspects of reconstruction precision, speed and success rate, etc. Based on the analysis of the advantages and disadvantages of the existing algorithms and according to the data characteristics of WSN, this paper proposes a more practical ITSAOMP algorithm with higher reconstruction accuracy and better stability and robustness. The experimental results show that the ITSAOMP algorithm can reconstruct the original signal with high probability and low distortion when the measurement matrix satisfies certain conditions. It combines compression sensing with periodic non-uniform sampling: periodic non-uniform sampling is a method to reduce sampling frequency and improve sampling precision effectively. It uses multi-channel sampling system to sample signals. According to the characteristics of non-uniform sampling system, the process of sampling and reconstruction is transformed into matrix or vector operation by using the theory of joint subspace, and the sparse signal reconstruction algorithm is applied to reconstruct the signal in the non-uniform sampling system with the help of CS theory. The simulation results show that the system not only realizes signal sampling and reconstruction, but also greatly reduces the sampling frequency and improves the reconstruction accuracy.
【学位授予单位】：南开大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN929.5;TP212.9

【参考文献】