一种改进的FCM聚类算法的混合矩阵估计
|
摘要
在增强信号稀疏性的基础上,对模糊C均值(fuzzy C-means, FCM)聚类算法进行改进,达到提高混合矩阵估计精度的目的,更好地解决欠定盲源分离问题。主要针对稀疏成分分析理论"两步法"中的混合矩阵估计算法改进,提出一种基于隶属度划分优化的FCM聚类算法。通过改变目标函数中的隶属度划分方式,来影响数据的归类,从而决定了混合矩阵中元素的估计精度。最后,将改进的算法用于语音信号仿真实验,完成混合矩阵估计。实验结果表明,用改进的算法所获得的矩阵估计误差小且精度高,可使归一化均方误差减小1.3 dB,角度偏差最多可减小1°。
The research is to improve the fuzzy C-means(FCM) clustering algorithm based on the enhancement of signal sparsity. Its goal is to achieve the purpose of improving the accuracy of mixing matrix estimation and solve the problem of underdetermined blind source separation better. In this paper, the mixing matrix estimation algorithm in the "two-step method" of sparse component analysis theory is improved mainly. After a brief description of the blind source separation problem, an FCM clustering algorithm based on membership classification optimization is proposed. The classification of the data is affected by changing the membership classification method in the objective function,determining the estimation accuracy of the elements in the mixing matrix. Finally, an improved algorithm was used in the speech signal simulation experiment to complete the mixing matrix estimation. The experimental results show that the matrix estimation error obtained by the improved algorithm is small, which can improve the estimation accuracy of the mixing matrix by reducing the normalized mean square error by 1.3 dB, and angular deviation by 1° at most.
The research is to improve the fuzzy C-means(FCM) clustering algorithm based on the enhancement of signal sparsity. Its goal is to achieve the purpose of improving the accuracy of mixing matrix estimation and solve the problem of underdetermined blind source separation better. In this paper, the mixing matrix estimation algorithm in the "two-step method" of sparse component analysis theory is improved mainly. After a brief description of the blind source separation problem, an FCM clustering algorithm based on membership classification optimization is proposed. The classification of the data is affected by changing the membership classification method in the objective function,determining the estimation accuracy of the elements in the mixing matrix. Finally, an improved algorithm was used in the speech signal simulation experiment to complete the mixing matrix estimation. The experimental results show that the matrix estimation error obtained by the improved algorithm is small, which can improve the estimation accuracy of the mixing matrix by reducing the normalized mean square error by 1.3 dB, and angular deviation by 1° at most.
引文
[1]余先川,胡丹.盲源分离理论与应用[M].北京:科学出版社,2011:3-9.
[2]UDDIN Z,AHMAD A,IQBAL M,et al.Applications of independent component analysis in wireless communication systems[J].Wireless personal communications,2015,83(4):2711-2737.
[3]COREY R M,SINGER A C.Underdetermined methods for multichannel audio enhancement with partial preservation of background sources[C]//2017 IEEE Workshop on Applications of Signal Processing To Audio and Acoustics.New Paltz,USA,2017:26-30.
[4]LI Chengjie,ZHU Lidong,LUO Zhongqiang.Underdetermined blind source separation of adjacent satellite interference based on sparseness[J].China communications,2017,14(4):140-149.
[5]李燕丽,吴士文,刘娅,等.基于FastICA盲源分离法去除土壤干扰的小麦生物量高光谱估算[J].生态学杂志,2017,36(4):1158-1164.
[6]BOBIN J,RAPIN J,LARUE A,et al.Sparsity and adaptivity for the blind separation of partially correlated sources[J].IEEE transactions on signal processing,2015,63(5):1199-1213.
[7]黄文威,凌云,徐敬成.基于FCM聚类和蚁群优化的WSN路由算法[J].信息与电脑(理论版),2018(1):38-40.
[8]张红丽.基于稀疏性增强的欠定盲源分离算法研究[D].秦皇岛:燕山大学,2016:30.
[9]DE LATHAUWER L,CASTAING J.Blind identification of underdetermined mixtures by simultaneous matrix diagonalization[J].IEEE transactions on signal processing,2008,56(3):1096-1105.
[10]朱然,李积英.几种优化FCM算法聚类中心的方法对比及仿真[J].计算机技术与发展,2015,25(5):17-20.
[11]刘锐,张宁.一种动态加权模糊聚类算法的研究[J].铁路计算机应用,2018,27(5):5-8,17.
[12]武玉坤.融合特征值与优化划分的改进FCM聚类算法[J].计算机工程与设计,2017,38(4):1076-1080.
[13]CHEN Peng,PENG Dezhong,ZHEN Liangli,et al.Underdetermined Blind Separation by Combining Sparsity and Independence of Sources[J].IEEE access,2017,5:21731-21742.
[2]UDDIN Z,AHMAD A,IQBAL M,et al.Applications of independent component analysis in wireless communication systems[J].Wireless personal communications,2015,83(4):2711-2737.
[3]COREY R M,SINGER A C.Underdetermined methods for multichannel audio enhancement with partial preservation of background sources[C]//2017 IEEE Workshop on Applications of Signal Processing To Audio and Acoustics.New Paltz,USA,2017:26-30.
[4]LI Chengjie,ZHU Lidong,LUO Zhongqiang.Underdetermined blind source separation of adjacent satellite interference based on sparseness[J].China communications,2017,14(4):140-149.
[5]李燕丽,吴士文,刘娅,等.基于FastICA盲源分离法去除土壤干扰的小麦生物量高光谱估算[J].生态学杂志,2017,36(4):1158-1164.
[6]BOBIN J,RAPIN J,LARUE A,et al.Sparsity and adaptivity for the blind separation of partially correlated sources[J].IEEE transactions on signal processing,2015,63(5):1199-1213.
[7]黄文威,凌云,徐敬成.基于FCM聚类和蚁群优化的WSN路由算法[J].信息与电脑(理论版),2018(1):38-40.
[8]张红丽.基于稀疏性增强的欠定盲源分离算法研究[D].秦皇岛:燕山大学,2016:30.
[9]DE LATHAUWER L,CASTAING J.Blind identification of underdetermined mixtures by simultaneous matrix diagonalization[J].IEEE transactions on signal processing,2008,56(3):1096-1105.
[10]朱然,李积英.几种优化FCM算法聚类中心的方法对比及仿真[J].计算机技术与发展,2015,25(5):17-20.
[11]刘锐,张宁.一种动态加权模糊聚类算法的研究[J].铁路计算机应用,2018,27(5):5-8,17.
[12]武玉坤.融合特征值与优化划分的改进FCM聚类算法[J].计算机工程与设计,2017,38(4):1076-1080.
[13]CHEN Peng,PENG Dezhong,ZHEN Liangli,et al.Underdetermined Blind Separation by Combining Sparsity and Independence of Sources[J].IEEE access,2017,5:21731-21742.