论文部分内容阅读
Bofill(2001)等人首次针对两个传感器的稀疏信号盲分离问题进行了讨论.但也正如Bofill自己所指出的那样,此方法存在局限性,特别是其中的势函数的参数选择缺乏理论指导,而且此方法不宜推广到具有三个或更多的传感器的情形.因此这里回避Bofill势函数方法,建立了K-PCA方法(即K-聚类与主成分分析PCA相结合的方法).新方法克服了Bofill方法参数选择的困难,可以方便地应用于三个及其以上传感器的情况,而且具有实现简单、混叠矩阵估计精度高的特点.另外,为了检验混叠矩阵A的估计是否一定有效,给出了相应的判别准则.仿真结果表明了该方法的可行性和准确性.
For the first time, Bofill et al. (2001) discussed the sparse signal blind separation problem of two sensors, but as Bofill himself pointed out, this method has its limitations. Especially, the choice of parameters in potential functions lacks theoretical guidance, And this method should not be generalized to have three or more sensors.Therefore, we avoid the Bofill potential function method and establish a K-PCA method (that is, a combination of K-clustering and principal component analysis PCA) .The new method Overcomes the difficulty of parameter selection of the Bofill method and can be conveniently applied to three or more sensors and has the advantages of simple implementation and high precision of the estimation of the aliasing matrix.In addition, in order to test whether the estimation of the aliasing matrix A is effective , The corresponding criterion is given.The simulation results show the feasibility and accuracy of this method.