一种改进的势函数欠定盲源分离算法

发布时间：2018-04-27 22:46

本文选题：欠定盲源分离 + 混合矩阵估计　；参考：《西安电子科技大学学报》2014年06期

【摘要】：针对原有的拉普拉斯混合模型势函数法复杂度高、随机选取部分观测数据点作为初始聚类中心的算法聚类结果不稳定、准确率低的问题,提出了一种改进的势函数欠定盲源分离算法.该算法在基于密度概念的基础上,以簇内距离小、簇间距离大为原则,选取部分高密度点作为势函数的初始聚类中心.理论分析与仿真实验表明,改进算法的复杂度大大降低,而估计准确度降低很少.在信噪比为10dB时,该算法仿真时间降为原始势函数法的5%;相对随机选取算法,在计算复杂度基本一致的前提下,该算法的估计准确度大大提高,源信号个数估计准确率由61%提高到85%,混合矩阵估计误差由0.47下降为0.27.
[Abstract]:In view of the high complexity of the original Laplacian hybrid model potential function method, the clustering results of random selected observation data points as the initial clustering center are unstable and the accuracy is low. An improved blind source separation algorithm with underdetermined potential function is proposed. Based on the concept of density, the algorithm selects some high-density points as the initial clustering center of the potential function based on the principle of small intra-cluster distance and large inter-cluster distance. Theoretical analysis and simulation experiments show that the complexity of the improved algorithm is greatly reduced, but the estimation accuracy is very low. When the SNR is 10dB, the simulation time of the algorithm is reduced to 5 parts of the original potential function method, and the estimation accuracy of the algorithm is greatly improved when the computational complexity is basically the same compared with the random selection algorithm. The accuracy of source number estimation is improved from 61% to 85%, and the error of mixed matrix estimation is reduced from 0.47 to 0.27.
【作者单位】：西安电子科技大学综合业务网理论及关键技术国家重点实验室;
【基金】：国家自然科学基金资助项目(61201134,61201135) 中央高校基本科研业务费专项资金资助项目(72124669) 高等学校学科创新引智计划资助项目(B08038) 重大专项基金资助项目(2012ZX03001027-001)
【分类号】：TN911.7

【共引文献】