针对混合矩阵估计算法中传统的噪声环境下基于密度的空间聚类(density-based spatial clustering of applications with noise，DBSCAN)算法需要人为设定邻域半径以及核心点数这一问题，提出双约束粒子群优化(double constrained particle swarm optimization，DCPSO)算法，对 DBSCAN 算法的邻域半径参数进行寻优，将得到的最优参数作为DBSCAN算法的参数输入，然后计算聚类中心，完成混合矩阵估计。针对基于距离排序的源信号数目估计算法存在依靠经验参数的选取且不具备噪声点剔除能力的问题，提出了最大距离排序算法。实验结果表明，所提算法较相应的对比算法皆有提升，源信号数目估计准确率较原算法提高近40％，混合矩阵估计的误差较对比算法提升3 dB以上，且所提算法在收敛速度上优于原算法。
Estimation of mixture matrix of density clustering algorithm based on improved particle swarm optimization algorithm
Aiming at the problem that the traditional density-based spatial clustering of applications with noise(DBSCAN)algorithm in the mixing matrix estimation algorithm needs to artificially set the neighborhood radius and the number of core points,a double constrained particle swarm optimization(DCPSO)algorithm is proposed.The neighborhood radius parameters of the DBSCAN algorithm are optimized,and the obtained optimal parameters are used as the parameter input of the DBSCAN algorithm,and then the clustering center is calculated to complete the mixing matrix estimation.Aiming at the problem that the source signal number estimation algorithm based on distance sorting relies on the selection of empirical parameters and does not have the ability to eliminate noise points,a maximum distance sorting algorithm is proposed.The experimental results show that the improved algorithm is improved.The accuracy of source signal number estimation is nearly 40％higher than that of the original algorithm.The error of mixing matrix estimation is more than 3 dB higher than that of the comparison algorithm.Moreover,the proposed algorithm has a better convergence speed than the original algorithm.
underdetermined blind source separationparticle swarm optimization(PSO)density space clusteringmixing matrix estimation
NETL
NSTL
万方数据
