计算机应用
計算機應用
계산궤응용
COMPUTER APPLICATION
2015年
z1期
331-334
,共4页
压缩感知%稀疏信号%贪婪算法%压缩采样匹配追踪%正交匹配追踪
壓縮感知%稀疏信號%貪婪算法%壓縮採樣匹配追蹤%正交匹配追蹤
압축감지%희소신호%탐람산법%압축채양필배추종%정교필배추종
Compressed Sensing(CS)%sparse signal%greedy algorithm%Compressed Sampling Matching Pursuit(CoSaMP)%Orthogonal Matching Pursuit(OMP)
针对压缩采样匹配追踪( CoSaMP)算法重构精度相对较差的问题,为了提高算法的重构性能,提出了一种基于伪逆处理改进的压缩采样匹配追踪( MCoSaMP)算法。首先,在迭代前,对观测矩阵进行伪逆处理,以此来降低原子间的相干性,从而提高原子选择的准确性;然后,结合正交匹配追踪算法( OMP),将OMP算法迭代K次后的原子和残差作为CoSaMP算法的输入;最后,每次迭代后,通过判断残差是否小于预设阈值来决定算法是否终止。实验结果表明,无论是对一维高斯随机信号还是二维图像信号,MCoSaMP算法的重构效果优于CoSaMP算法,能够在观测值相对较少的情况下,实现信号的精确重构。
針對壓縮採樣匹配追蹤( CoSaMP)算法重構精度相對較差的問題,為瞭提高算法的重構性能,提齣瞭一種基于偽逆處理改進的壓縮採樣匹配追蹤( MCoSaMP)算法。首先,在迭代前,對觀測矩陣進行偽逆處理,以此來降低原子間的相榦性,從而提高原子選擇的準確性;然後,結閤正交匹配追蹤算法( OMP),將OMP算法迭代K次後的原子和殘差作為CoSaMP算法的輸入;最後,每次迭代後,通過判斷殘差是否小于預設閾值來決定算法是否終止。實驗結果錶明,無論是對一維高斯隨機信號還是二維圖像信號,MCoSaMP算法的重構效果優于CoSaMP算法,能夠在觀測值相對較少的情況下,實現信號的精確重構。
침대압축채양필배추종( CoSaMP)산법중구정도상대교차적문제,위료제고산법적중구성능,제출료일충기우위역처리개진적압축채양필배추종( MCoSaMP)산법。수선,재질대전,대관측구진진행위역처리,이차래강저원자간적상간성,종이제고원자선택적준학성;연후,결합정교필배추종산법( OMP),장OMP산법질대K차후적원자화잔차작위CoSaMP산법적수입;최후,매차질대후,통과판단잔차시부소우예설역치래결정산법시부종지。실험결과표명,무론시대일유고사수궤신호환시이유도상신호,MCoSaMP산법적중구효과우우CoSaMP산법,능구재관측치상대교소적정황하,실현신호적정학중구。
Aiming at the problem that Compressed Sampling Matching Pursuit ( CoSaMP) algorithm has low accuracy in reconstruction, in order to improve the reconstruction performance of CoSaMP algorithm, , based on pseudo-inverse processing, an improved greedy algorithm—Modified Compressed Sampling Matching Pursuit ( MCoSaMP ) was proposed. Firstly, before each iteration, the proposed algorithm did pseudo-inverse processing on observation matrix, which could reduce the coherence between the atoms, thereby improving the accuracy of the selected atoms. Secondly, combined with Orthogonal Matching Pursuit ( OMP) algorithm, MCoSaMP used the atoms and residual as the input parameters of CoSaMP after OMP algorithm iterating K times. Finally, after each iteration, the residual was used to determine whether to stop algorithm by being under a preset threshold or not. The experimental results show that the proposed algorithm performs better than CoSaMP algorithm for both one-dimensional Gaussian random signal and two-dimensional image signal, which can exactly reconstruct the original signal with relatively small number of observations.