摘要
欠定盲源分离问题中,针对传统FCM算法(fuzzy C-means,FCM)需要预先设定聚类数目和初始聚类中心,以及聚类结果易受噪点干扰的问题,提出一种基于密度结构分析的改进FCM聚类算法,并利用改进后的算法实现混合矩阵估计。这一改进算法首先用OPTICS(ordering points to identify the clustering structure,OPTICS)算法对信号进行密度结构分析,得到能反映信号密度结构的可达距离序列,从中确定出初始聚类中心和聚类数目,实现对FCM初始参数优化;而后进一步将可达序列作为动态加权因子应用到FCM目标函数中,实现对目标函数的优化。仿真结果表明,本文提出的改进算法可以从初始参数和目标函数2方面实现对传统FCM算法的优化,提高聚类的稳定性和最终混合矩阵的鲁棒性。
In the problem of underdetermined blind source separation,traditional fuzzy c-means(FCM)algorithm needs to set the number of clusters and the initial value in advance,and is easy to be disturbed by noise.To solve this problem,an improved FCM clustering algorithm based on density structure analysis is proposed,realizing the hybrid matrix estimation.The ordering points to identify the clustering structure(OPTICS)algorithm is firstly used to analyze the density structure of signals,and the reachable distance sequence reflecting the density structure of signal is obtained,and then the initial clustering center and the number of clusters are searched to optimize the initial parameters of FCM.Then,the reachable sequence is further applied to the objective function of FCM as a dynamic weighting factor to optimize the objective function.Simulation results show that the improved algorithm can optimize the traditional FCM algorithm from the initial parameters and objective functions,and improve the stability of clustering and the robustness of the final mixed matrix.
作者
刘阳
高敬鹏
LIU Yang;GAO Jingpeng(College of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)
出处
《应用科技》
CAS
2021年第3期57-63,共7页
Applied Science and Technology
关键词
欠定盲源分离
稀疏成分分析
混合矩阵估计
FCM
OPTICS
数据聚类
可达距离
动态加权因子
underdetermined blind source separation
sparse component analysis
mixed matrix estimation
FCM
OPTICS
data clustering
reachable distance
dynamic weighting factor
