摘要: |
在压缩感知(CS)技术中,基于最小洛伦兹范数的重构算法在脉冲噪声环境下具有较强的鲁棒性。但是,洛伦兹硬阈值追踪(LHTP)算法采用残差的负梯度更新信号的估计量,从中引入了噪声的影响。基于此,提出了洛伦兹双阈值余量追踪(LDTRP)算法,通过迭代估计的重构误差来修正信号的估计量,并利用残差设置新的野点阈值。实验结果表明,无论信号是否受到噪声影响,LDTRP算法均能获得更好的信号重构质量。另外,在含有脉冲噪声的环境中,相比洛伦兹迭代硬阈值(LIHT)算法与LHTP算法,所提算法降低了对脉冲数量以及脉冲功率变化的敏感度,提高了重构性能。 |
关键词: 压缩感知 信号重构 脉冲噪声 洛伦兹范数 |
DOI: |
|
基金项目:国家科技重大专项(2016ZX03002010-003) |
|
Lorentzian double threshold residual pursuit signal reconstruction with impulsive noise |
DUAN Hongguang,LI Lingxin |
(a.School of Communications and Information Engineering;b.Communication Core Chip,Protocols and Application Innovation Team,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;School of Communications and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China) |
Abstract: |
In compressed sensing(CS) technology,a reconstruction algorithm based on the minimum Lorentzian norm has strong robustness in impulsive noise environment.However,the negative gradient of residual is used to update the estimator of signal in the the Lorentzian Hard Threshold Pursuit(LHTP) algorithm,which introduces noise in the estimator.For the defect,the Lorentzian Double Threshold Residual Pursuit(LDTRP) algorithm is proposed in which the estimator of signal is corrected by estimating the reconstructed error in the iteration and the residual is used to set the new outliers threshold.Experimental results show that the LDTRP algorithm can achieve better signal recovery quality regardless of whether the signal contains noise or not.In addition,in the environment with impulsive noise,compared with those of the Lorentzian Iterative Hard Thresholding(LIHT) and LHTP algorithms,the sensitivity of the LDTRP algorithm to impulse number and impulse power change are reduced,and the reconstruction performance is improved. |
Key words: compressed sensing signal reconstruction impulsive noise Lorentzian norm |