基于改进源信号数目估计算法的欠定盲分离

发布时间：2019-07-27 13:18

【摘要】：两步法是解决稀疏信号欠定盲分离的一种常用方法,通常首先利用K-means聚类算法估计混叠矩阵,然后利用最短路径法恢复源信号。在使用K-means聚类算法时要求知道源信号的数目,而现实中往往不知道源信号的数目,需要对其进行估计。因此研究了聚类有效性评价指标——BWP指标,结合粒子群算法,提出了一种改进的确定源信号数目的算法,并将这种算法引入到欠定盲分离。实验表明,提出的算法在保证分离精度的同时能缩短分离时间,并可节省一定的内存,在观测信号数据量大时,这种优势更加明显。
[Abstract]:Two-step method is a common method to solve the underdetermined blind separation of sparse signals. K-means clustering algorithm is usually used to estimate the aliasing matrix, and then the shortest path method is used to recover the source signal. When using K-means clustering algorithm, we need to know the number of source signals, but in reality, we often do not know the number of source signals, so we need to estimate them. Therefore, the BWP index, an evaluation index of clustering validity, is studied. Combined with particle swarm optimization (Particle Swarm Optimization), an improved algorithm for determining the number of source signals is proposed, and this algorithm is introduced into underdetermined blind separation. The experimental results show that the proposed algorithm can not only ensure the separation accuracy, but also shorten the separation time and save a certain amount of memory, which is more obvious when the amount of observation signal data is large.
【作者单位】：湖南大学电气与信息工程学院;合肥工业大学电气与自动化工程学院;中国电子科技集团公司第41研究所电子测试技术国防科技重点实验室;
【基金】：国家杰出青年科学基金资助项目(50925727) 国家自然科学基金资助项目(60876022,61102039,51107034) 湖南省科技计划资助项目(2011J4,2011JK2023) 湖南省自然科学基金资助项目(12JJA004) 国防计划资助项目(C1120110004,9140A27020211DZ5102) 国家教育部科学技术研究重大项目(313018) 中央高校基本科研业务费计划资助项目
【分类号】：TN911.7

【参考文献】