基于基底合并的分片稀疏恢复

如我们所知,诸如视频和图像等信号可以在某些框架下被表示为稀疏信号,因此稀疏恢复(或稀疏表示)是信号处理、图像处理、计算机视觉、机器学习等领域中被广泛研究的问题之一.通常大多数在稀疏恢复中的有效快速算法都是基于求解$l^0$或者$l^1$优化问题.但是,对于求解$l^0$或者$l^1$优化问题以及相关算法所得到的理论充分性条件对信号的稀疏性要求过严.考虑到在很多实际应用中,信号是具有一定结构的,也即,信号的非零元素具有一定的分布特点.在本文中,我们研究分片稀疏恢复的唯一性条件和可行性条件.分片稀疏性是指一个稀疏信号由多个稀疏的子信号合并所得.相应的采样矩阵是由多个基底合并组成.考虑到采样矩阵的分块结构,我们引入了子矩阵的互相干性,由此可以得到相应$l^0$或者$l^1$优化问题可精确恢复解的稀疏度的新上界.本文结果表明.通过引入采样矩阵的分块结构信息.可以改进分片稀疏恢复的充分性条件.以及相应$l^0$或者$l^1$优化问题整体稀疏解的可靠性条件.     

压缩感知和稀疏优化简介

介绍压缩感知和稀疏优化的基本概念、理论基础和算法概要. 压缩感知利用原始信号的稀疏性,从远少于信号元素个数的测量出发,通过求解稀疏优化问题来恢复完整的原始稀疏信号. 通过一个小例子展示这一过程,并以此说明压缩感知和稀疏优化的基本理念. 接着简要介绍用以保证l₁凸优化恢复稀疏信号的零空间性质和RIP条件. 最后介绍求解稀疏优化的几个经典算法.     

基于l2/lq(q=2/3)最小化模型的块稀疏信号恢复

该文主要研究了块稀疏信号的恢复问题.利用q块限制等距性质(0<q≤1),通过极小化混合l₂/l_q(q=2/3)范数,建立了块稀疏信号恢复的一个充分条件,并且得到了在有噪声情形下信号恢复的误差界.通过数值实验,验证了该模型对于块稀疏信号的恢复有较高的成功率.     

无约束的l2,1-分析法重构冗余紧框架下分块稀疏信号的条件

文章主要利用分块稀疏信号的凸分解技术分析无约束的l_2,1-分析模型,建立无约束的l_2,1-分析法重构冗余紧框架下分块稀疏信号的条件,其条件基于紧框架下的限制等距性质.首先,利用分块稀疏信号的凸分解技术建立两个重要技术引理.其次,基于发展的两个技术引理建立无约束的l_2,1-分析法恢复冗余紧框架下分块稀疏信号新的恢复条件,其条件基于紧框架下的限制等距性质,改进了现存最好的恢复条件.最后,设计数值实验,说明无约束的l_2,1-分析法重构冗余紧框架下分块稀疏信号的性能.     

预估校正法在大规模信号重构问题中的应用

本文研究预估校正法在大规模信号重构中的应用问题.利用预估校正方法解?_1正则化最小二乘问题,获得了理想的信号恢复效果.数值实验表明提出的算法对于解决大规模稀疏信号恢复问题是有效的.     

基于最小化似零伪范数的稀疏信号恢复算法

针对欠定系统中出现的稀疏信号恢复问题,提出了一种基于最小化近似零伪范数的处理方法,算法首先结合反正切函数构造出代价函数,再融合最速下降法和扩展牛顿迭代法逐步迭代寻优,并给出了算法的收敛性分析,数值仿真实验结果表明,与经典的稀疏信号恢复算法相比,方法有更好的计算速度和恢复精度.     

利用分片贪婪算法求解分片稀疏恢复问题

在实际应用中,有一些信号是具有分片的结构的.本文我们提出一种分片正交匹配追踪算法(P\_OMP)来求解分片稀疏恢复问题,旨在保护分片信号中的分片结构(或者小尺度非零元).P\_OMP算法是基于CoSaMP和OMMP算法的思想上延伸出的一种针对分片稀疏问题的贪婪算法. P\_OMP算法不仅仅具有OMP算法的优势,还能够在比CoSaMP方法更松弛的条件下得到同样的误差下降速率.进一步,P\_OMP~算法在保护分片稀疏信号的尺度细节信息上表现的更好.数值实验表明相比于CoSaMP, OMP, OMMP和BP算法, P\_OMP算法在分片稀疏恢复上更有效更稳定.     

多面体块稀疏表示和非凸压缩感知

压缩感知(compressed sensing,CS)理论表明稀疏信号可以从欠定系统中被准确恢复,但在很多实际应用中,信号不一定有标准稀疏性而可能拥有一些其他的结构特点,典型的一种就是块稀疏信号,它的非零元仅在很少的一些块中出现.本文考虑从很少的线性测量中恢复块稀疏信号,并得到经混合l2/lq(0＜q≤1)最小化准确重...     

不确定量测联合信号重构的稳定性

研究了分布式压缩感知(Distributed Compressed Sensing,DCS)理论对联合稀疏信号进行联合重构的稳定性问题.文中讨论的联合稀疏信号模型中含有两个近似稀疏信号,且信号的量测过程中带有噪声.证明利用分布式压缩感知思想对近似稀疏的联合稀疏信号的联合稀疏重构具有稳定性,刻画了重构信号的误差,并与单个信号的稀疏重构导致的误差进行了比较,证明了在一定条件下,利用分布式压缩感知思想对信号进行联合重构的误差界小于单个信号重构的误差界.     

求解稀疏相位恢复问题的随机交替方向法(英文)

近年来稀疏相位恢复问题受到了越来越多的关注.本文提出了一种随机交替方法方法求解稀疏相位恢复问题,该算法采用硬阈值追踪算法求解带稀疏约束的最小二乘子问题.大量的数值实验表明,该算法可以通过O(s log n)次测量(理论上最少测量值)稳定的恢复n维s稀疏向量,并且在随机初值下可以获得全局收敛性.     

Band-limited signal extrapolation by truncated Bernstein polynomials

An algorithm is given for everywhere extrapolating a band-limited signal known only on an interval of arbitrary finite length. The scheme utilizes a finite number of equally spaced samples of the given function and provides a time-limited polynomial approximation. The approximation functions are shown to converge everywhere pointwise and uniformly in any compact interval to the band-limited signal. When the original band-limited signal is also Lebesgue integrable it is also established that the Fourier transform of the approximating signal converges uniformly to the Fourier transform of the original signal.     

A centrosymmetric matrix based technique for the interpolation of a Hermitian signal

This work is concerned with the signal interpolation problem, i.e. given only samples of a signal, a method is derived for evaluating its samples on finer grids. The derivation is based on a discrete‐time decimation formula. In the special case where the known samples have the Hermitian property, two schemes are presented and mathematically proved to result in interpolated points having the same property. The first scheme does not utilize the known sample at the origin and results in a square system of equations to be solved for the unknown interpolated points of the signal. The second scheme has the merit of utilizing all the known samples, but it results in an overdetermined system of equations to be solved by the least squares method. The exploitation of the elegant properties of the involved centrosymmetric matrices is central to the treatment presented here. Copyright © 2002 John Wiley & Sons, Ltd.     

Signal modeling using the gradient search

This letter presents a new method for continuous signal modeling. Firstly, the continuous signal can be represented as a function of the trigonometric functional extension (Fourier series). Fourier series of the signal are parameterized by the fundamental frequency and unknown parameters. Then, the gradient-based iterative identification algorithm is derived, for estimating parameters of the signal model with known and unknown frequencies, separately. Finally, the simulation results indicate that the proposed algorithm is effective.     

Vibrational resonance in a time-delayed genetic toggle switch

Biological oscillators can respond in a surprising way when they are perturbed by two external periodic forcing signals of very different frequencies. The response of the system to a low-frequency signal can be enhanced or depressed when a high-frequency signal is acting. This is what is known as vibrational resonance (VR). Here we study this phenomenon in a simple time-delayed genetic toggle switch, which is a synthetic gene-regulatory network. We have found out how the low-frequency signal changes the range of the response, while the high-frequency signal influences the amplitude at which the resonance occurs. The delay of the toggle switch has also a strong effect on the resonance since it can also induce autonomous oscillations.     

Local sampling set conditions in weighted shift-invariant signal spaces

How to represent a continuous signal in terms of a discrete sequence is a fundamental problem in sampling theory. Most of the known results concern global sampling in shift-invariant signal spaces. But in fact, the local reconstruction from local samples is one of the most desirable properties for many applications in signal processing, e.g. for implementing real-time reconstruction numerically. However, the local reconstruction problem has not been given much attention. In this article, we find conditions on a finite sampling set X such that at least in principle a continuous signal on a finite interval is uniquely and stably determined by their sampling value on the finite sampling set X in shift-invariant signal spaces.     

Complete signal processing bases and the Jacobi group

The continuous windowed Fourier and wavelet transforms are created from the actions of the Heisenberg and affine groups, respectively. Both wavelet and windowed Fourier bases are known to be complete; that is, the only signal which is orthogonal to every element of each basis is the zero signal. The Jacobi group is a group which contains both the Heisenberg and affine groups, and it can also be used to produce bases for signal processing. This paper investigates completeness for bases of one and two real variables which are produced by the Jacobi group.     

Design of RLS Wiener fixed-lag smoother using covariance information in linear discrete stochastic systems

In this paper, we propose a new design for the recursive least-squares (RLS) Wiener fixed-lag smoother and filter in linear discrete-time wide-sense stationary stochastic systems. It is assumed that the signal is observed with additive white observation noise. The signal is uncorrelated with the observation noise. The estimators require knowledge of the system matrix, the observation matrix and the variance of the state vector. These quantities can be obtained from the auto-covariance function of the signal. In the estimation algorithms, moreover, the variance of the observation noise is assumed to be known, as a priori information.     

Signal estimation from incomplete data

We give explicit analytical formulas for finding a signal with the known compact support from its spectrum known on a finite frequency band.     

On two estimates related to the change-point problem

We consider the problem of estimating a smooth functional of an unknown signal with discontinuity from Gaussian observations. The signal is a known function depending on an unknown parameter. This problem is closely related to the famous change-point problem. We obtain an asymptotic likelihood ratio process for the noise level tending to 0. Bayesian and maximum likelihood estimates are constructed and their relative efficiency is studied. Some simulation results and conclusions on non-asymptotic behavior of these estimates are presented.     

相关技术在滚动轴承故障诊断中的应用

众所周知自相关技术能够有效地从随机嘈声中提取周期性信号.在滚动轴承元件的故障诊断中,由于信号图象不同,我们无法从原信号直接得到故障的情况,这种信号的显示就像宽频带随机嘈声在自相关函数的显示一样.本文中,信号经过了预处理,其结果证明是行之有效的.应用自相关技术我们还可以得到已测定的可比采样,这对于建立轴承运行条件的数据基础和进行故障诊断是十分重要的.