基于复Givens矩阵与蝙蝠优化的卷积盲分离算法

doi:10.11959/j.issn.1000-436x.2016138

摘要/Abstract

摘要：

针对传统卷积混合盲分离待求参数多、分离效果易受分离矩阵初值影响的局限性，提出了基于复Givens矩阵与蝙蝠优化的频域求解算法。算法采用复Givens矩阵表示分离矩阵，减少了待求参数，降低了求解难度和计算量。利用蝙蝠算法代替梯度算法优化求解旋转角度完成各频点线性瞬时混合复信号的盲分离，全局收敛性更强。此外，由于对源信号的先验知识要求较少，可以分离服从多种分布的信号。仿真实验表明，该算法可有效地实现卷积混合盲分离。

关键词: 盲分离, 卷积混合, 蝙蝠算法, 复Givens矩阵

Abstract:

For the limitations such as many unknown parameters, the separation accuracy was easily influenced by initial value of separation matrix in traditional convolutive blind separation, a kind of frequency method based on complex Givens matrix and bat optimization was proposed. The algorithm used a series of complex Givens matrices to denote separation matrix, it reduced unknown parameters, decreased the difficulty and the amount of calculations as a result. Be-sides, the algorithm utilized bat algorithm instead of conventional gradient algorithm to optimize rotation angles and completed the separation of complex linear instantaneous mixing signals at each frequency point, the use of bat algorithm made the optimization ability better. In addition, little prior information was needed and signals following various distri-butions could be separated. Simulation results show that the proposed method can realize convolutive blind separation ef-ficiently.

Key words: blind separation, convolutive mixtures, bat algorithm, complex Givens matrix

贾志成,韩大伟,陈雷,郭艳菊,许浩达. 基于复Givens矩阵与蝙蝠优化的卷积盲分离算法[J]. 通信学报, 2016, 37(7): 107-117.

Zhi-cheng JIA,Da-wei HAN,Lei CHEN,Yan-ju GUO,Hao-da XU. Convolutive blind separation algorithm based on complex Givens matrix and bat optimization[J]. Journal on Communications, 2016, 37(7): 107-117.

图/表 8

图1

图2

表1

图3

表2

采用不同算法用于3路卷积混合信号盲分离所得相关系数绝对值及重构信噪比"

算法		相关系数绝对值				重构信噪比 $\frac{S}{N}$ /dB
算法	第1路	第2路	第3路	平均值	第1路	第2路	第3路	平均值
Fast-ICA算法	0.820 6	0.802 6	0.850 9	0.824 7	2.052 3	0.679 8	2.628 9	1.787 0
IVA算法	0.960 4	0.951 2	0.937 1	0.949 5	12.344 7	6.155 9	8.920 7	9.140 4
C-FastICA算法	0.973 1	0.943 5	0.923 4	0.946 6	13.089 6	5.760 6	7.426 2	8.758 8
KM-G算法	0.974 4	0.978 6	0.986 1	0.979 7	16.382 9	13.031 2	16.226 1	15.213 4
JADE算法	0.978 7	0.976 5	0.952 3	0.969 1	12.551 1	7.880 3	6.942 0	9.124 5
本文算法	0.976 7	0.981 2	0.987 2	0.981 7	16.825 3	13.881 6	15.938 6	15.548 5

表2

图4

表3

表4

参考文献 26

[1]	YANG Z Y , XIANG Y , RONG Y , et al. A convex geometry-based blind source separation method for separating nonnegtive sources[J]. IEEE Transactions on Neural Networks and Learning Systems, 2015,26(8):1635-1644.
[2]	RIVET B . Source separation of multimodal data: a second-order ap-proach based on a constrained joint block decomposition of covariance matrices[J]. IEEE Signal Processing Letters, 2015,22(6):681-685.
[3]	FU G S , PHLYPO R , ANDERSON M , et al. Blind source separation by entropy rate minimization[J]. IEEE Transactions on Signal Processing, 2014,62(16):4245-4255.
[4]	OUEDRAOGO W S B , SOULOUMIAC A , JAIDANE M , et al. Non-negative blind source separation algorithm based on minimum aperture simplicial cone[J]. IEEE Transactions on Signal Processing, 2014,62(2):376-389.
[5]	WON Y G , LEE S Y . Convolutive blind signal separation by estimat-ing mixing channels in time domain[J]. Electronics Letters, 2008,44(21):1277-1279.
[6]	CASTELLA M , RHIOUI S , MOREAU E , et al. Quadratic higher order criteria for iterative blind separation of a MIMO convolutive mixture of sources[J]. IEEE Transactions on Signal Processing, 2007,55(1):218-232.
[7]	BINGHAM E , HYVARIEN A . A fast fixed-point algorithm for inde-pendent component analysis of complex valued signals[J]. International Journal of Neural Systems, 2000,10(1):1-15.
[8]	LEE I , KIM T , LEE T W . Fast fixed-point independent vector analysis algorithms for convolutive blind source separation[J]. Signal Processing, 2007,87(8):1859-1871.
[9]	CARDOSO J F . Equivariant adaptive source separation[J]. IEEE Transactions on Signal Processing, 1996,44(12):3017-3029.
[10]	LI H , ADALI T . A class of complex ICA algorithms based on the kurtosis cost function[J]. IEEE Transactions on Neural Networks, 2008,19(3):408-420.
[11]	YANG X S . A new metaheuristic bat-inspired algorithm[C]// International Workshop on Nature Inspired Cooperative Strategies for Optimization. Granada, c2010:65-74.
[12]	PREMKUMAR K , MANIKANDAN B V . Speed control of brushless DC motor using bat algorithm optimized adaptive neuro-fuzzy infer-ence system[J]. Applied Soft Computing, 2015,32:403-419.
[13]	NIKNAM T , BAVAFA F , AZIZIPANAH-ABARGHOOEE R . New self-adaptive bat-inspired algorithm for unit commitment problem[J]. IET Science Measurement ＆ Technology, 2014,8(6):505-517.
[14]	JADDI N S , ABDULLAH S , HAMDAN A R . Optimization of neural network model using modified bat-inspired algorithm[J]. Applied Soft Computing, 2015,37:71-86.
[15]	覃和仁，谢胜利．基于QR分解与罚函数方法的盲分离算法[J]．计算机工程， 2003,29(17):55-57. QIN H R , XIE S L . Blind separation algorithm based on QR decom-position and penalty function[J]. Computer Engineering, 2003,29(17):55-57.
[16]	杜鹃，冯思臣．复矩阵的Givens变换及其QR分解[J]．成都理工大学学报(自然科学版)， 2011,38(6):693-695. DU J , FENG S C . Givens transformation and QR factorization of complex matrix[J]. Journal of Chengdu University of Technology(Science＆Technology Edition), 2011,38(6):693-695.
[17]	MATSUOKA K , NAKASHIMA S . Minimal distortion principle for blind source separation[C]// International Workshop on Independent Component Analysis and Blind Signal Separation. Osaka, c2001:722-727.
[18]	CARDOSO J F , SOULOUMIAC A . Blind beamforming for non-Gaussian signals[J]. IEE Proceedings F-Radar and Signal Processing, 1993,140(6):362-370.
[19]	陈晓军，成昊，唐斌．基于ICA的雷达信号欠定盲分离算法[J]．电子与信息学报， 2010,32(4):919-924. CHEN X J , CHENG H , TANG B . Underdetermined blind radar signal separation based on ICA[J]. Journal of Electronics ＆ Information Technology, 2010,32(4):919-924.
[20]	HYVARINEN A . Fast and robust fixed-point algorithms for inde-pendent component analysis[J]. IEEE Transactions on Neural Net-works, 1999,10(3):626-634.
[21]	陈雷，张立毅，郭艳菊，等．基于时间可预测性的差分搜索盲信号分离算法[J]．通信学报， 2014,35(6):117-125. CHEN L , ZHANG L Y , GUO Y J , et al. Blind signal separation algo-rithm based on temporal predictability and differential search algo-rithm[J]. Journal on Communications, 2014,35(6):117-125.
[22]	CICHOCKI A , AMARI S . Adaptive blind signal and image processing:learning algorithms and applications[M]. New York: Wiley, 2002.
[23]	DALI N , BOUAMAMA S . GPU-PSO: parallel particle swarm opti-mization approaches on graphical unit for constraint reasoning: case of max-CSPs[C]// International Conference on Knowledge Based on In-telligent Information and Engineering System. Singapore, c2002:1070-1080.
[24]	OUYANG A , Tang Z , ZHOU X , et al. Parallel hybrid PSO with CUDA for 1D heat conduction equation[J]. Computer ＆ Fluids, 2015,110:198-210.
[25]	TOUTOUH J , ALBA E . Parallel swarm intelligence for VANETs optimization[C]// International Conferrence on P2P, Parallel, Grid, Cloud and Internet Computing. Victoria, Spain, c2012:285-290.
[26]	SILVA E H M , BASTOS FILHO C J A . PSO efficient implementation on GPUs using low latency memory[J]. IEEE Latin America Transactions, 2015,13(5):1619-1624.

算法		相关系数的模		平均值
算法	第1路	第2路	第3路	平均值
KM-G	0.999 470 11	0.999 727 00	0.999 294 62	0.999 497 24
JADE	0.998 412 29	0.999 956 68	0.994 482 09	0.997 617 02
本文	0.999 470 34	0.999 727 09	0.999 294 91	0.999 497 45