摘要
针对时域盲解卷积存在求解变量多、收敛速度慢、容易陷入局部最优等问题进行了研究,提出一种防止遗传算法局部收敛的"监测策略",可以实时监控算法向全局最优解靠近的情况;同时对交叉概率、变异概率等关键技术进行相应设计,该算法能够自动跳出局部最优,快速地收敛于全局最优解。在概率密度估计的基础上,得到时域盲解卷积的基于最小互信息的分离准则。以此最小互信息准则确定遗传算法的寻优标准,快速地实现了时域盲解卷积。使用Matlab软件仿真验证了该时域盲解卷积算法的有效性。
Concerning the multi-variable solution, slow convergence and easily falling into a local optimum in time-domain blind deconvolution, a ＂ monitoring strategy＂ was proposed in order to prevent from local convergence of genetic algorithm. At the same time, some of the key technologies of the general genetic algorithm, such as crossover probability and mutation probability, were designed correspondingly so that the algorithm could automatically jump out of the local optimum solution, and rapidly converge in the global optimum solution. Separation criteria based on minimum mutual information of time-domain blind deconvolution was obtained on the basis of probability density estimation. That separation criterion used as algorithm optimization standard of genetic algorithm, time-domain blind deconvolution was realized quickly. By using Matlab software to simulate, the effectiveness of time-domain algorithm proposed is confirmed.
出处
《计算机应用》
CSCD
北大核心
2009年第5期1257-1260,共4页
journal of Computer Applications
基金
国家863计划项目(2002AA632080)
吉林省自然科学基金资助项目(20050705-6)
关键词
盲分离
盲卷积
遗传算法
最小互信息
概率密度估计
blind source separation
blind convolution
genetic algorithm
minimum mutual information
probability density estimation
