中文摘要：

针对压缩感知中测量矩阵性能判据混乱的问题,提出基于不同类型重构算法的测量矩阵性能判据.该判据比传统方法的可操作性更好,在矩阵的其他性质都近乎一致的条件下实现了对具有不同列不相关性矩阵的严格对比实验.实验结果表明,对于OMP（orthogonal matching pursuit）算法,测量矩阵的列不相关性可作为测量矩阵性能判据,但当列不相关性达到一定程度（即μcmax〉0.25）后对测量矩阵性能再无影响.对于BP（Basis Pursuit）算法,具有相同解空间的测量矩阵性能与列不相关性无关;测量矩阵的列不相关性不是判断测量矩阵性能的主要判据.一个矩阵能达到的理论上限就是等规模的哈达玛矩阵的重构能力.

英文摘要：

Aiming at the confusing problem of measurement matrix performance criteria, the performance crite-ria of the measurement matrix based on different types of reconstruction algorithms were proposed. The operate- bility of this criterion was better than the traditional method. The strict contrast experiments were implemented on matrices with different columns correlation by this method when other properties of matrices were almost i-dentical. Experimental results showed that column incoherence of a measurement matrix could be taken as the performance criteria of the measurement matrix for OMP （orthogonal matching pursuit）. However, the per-formance of the measurement matrix did not be affected when the correlation was enough between the columns.The performance of the measurement matrices with the same solution space had no relation with the column cor-relation for BP（Basis Pursuit）. Measurement matrix correlation was not the main criteria to determine the per-formance of the measurement matrix. Theoretical upper limit of the matrix was reconstruction ability of Had- amard matrix in same scale.