压缩感知的稀疏字典学习在信号重建中的应用

Application of Compressed Sensing Sparse Dictionary Learning in Signal Reconstruction

针对时域信号在压缩感知的正交基下重建存在较大误差的问题,提出了一种字典学习方法稀疏表示连续时域信号。该方法基于K-奇异值分解(K-SVD)字典更新学习算法对原始数据稀疏化,在压缩感知理论下,通过观测矩阵得到的少量测量值利用正交匹配跟踪算法重新构建信号的非零元素系数矩阵,从而达到信号的近乎完美的重建。仿真实验结果表明,字典学习比传统正交基在信号重建上有更好的效果。

A dictionary learning method is proposed to sparse represent continuous time-domain signals in this paper.This method is based on k-svd dictionary update learning algorithm to sparse the original data.Under the compression perception theory,a small amount of measured values obtained by observation matrix are reconstructed using orthogonal matching tracking algorithm to reconstruct the non-zero element coefficient matrix of the signal,so as to achieve the nearly perfect reconstruction of the signal.