摘要
压缩感知是通过对信号信息采样的信号处理新方法,它对可压缩信号可以大大降低采样数据。为提高噪声信号在压缩感知中的重构精度,本文提出了一种基于对观测矩阵奇异值分解的噪声信号重构算法,该算法对随机观测矩阵进行奇异值分解,通过均值算法修改对角矩阵的特征值,产生新的观测矩阵用于线性测量,理论证明了新观测矩阵比原观测矩阵具有更高的重构精度。仿真结果表明,算法重构精度在一维信号提高了3%～5%,二维信号的PSNR值提高1～2 dB。
The novel theory of compressed sensing （CS） reduces the samples of compressible signal sharply by infor- mation sampling. In order to improve reconstruction accuracy of noise signal for CS, a singular value decomposition （SVD） noise signal reconstruction algorithm is presented. This algorithm decomposes the random observation matrix by SVD, modifies the diagonal matrix eigenvalues by mean algorithm and obtains the new linear measurement matrix. The theorem proves that the new matrix owns higher reconstruction accuracy than the original matrix. Simulation resuits show that this paper＇s algorithm can promote reconstruction accuracy 3% -5% for one dimensional signal and 1 -2 dB for two dimensional signal.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2012年第12期2655-2660,共6页
Chinese Journal of Scientific Instrument
基金
国家杰出青年科学基金(50925727)
国家自然科学基金(60876022)资助项目
