动量项盲源分离算法及其性能优化策略

发布时间：2018-06-07 01:54

本文选题：盲源分离 + 动量项　；参考：《电子学报》2014年01期

【摘要】：动量项技术是用来改善自适应盲源分离算法分离性能的有效手段,但算法在融入动量项后,其收敛特性对于动量因子的选取数值较为敏感,且算法的稳态性能仍要受到步长参数的限定.本文首先给出了动量项盲源分离算法的设计原理,分析了现有算法存在的两个缺陷性问题;然后利用梯度下降法构造了具有在线调整特性的动量因子自适应迭代规则,通过对动量因子的实时更新以消除固定动量因子算法的性能缺陷;在此基础上,基于凸组合理论设计了不同步长参数下两个变动量因子算法的自适应优化组合方案,从而在一定程度上缓解了步长参数对于算法性能的限定.在不同环境下进行的仿真实验表明,本文针对动量项盲源分离算法所设计的优化策略能够有效消除其所存在的缺陷问题,在确保快速收敛的同时,又能获取较小的稳态误差.
[Abstract]:Momentum term technique is an effective method to improve the separation performance of adaptive blind source separation algorithm, but when the momentum term is incorporated, the convergence property of the algorithm is sensitive to the selection of momentum factor. The steady-state performance of the algorithm is still limited by step parameters. In this paper, the design principle of the blind source separation algorithm for momentum term is given, and two defects of the existing algorithms are analyzed, and then the adaptive iterative rule of momentum factor with on-line adjustment is constructed by using gradient descent method. By updating the momentum factor in real time, the performance defect of the fixed momentum factor algorithm is eliminated, and based on the convex combination theory, the adaptive optimal combination scheme of two variable factor algorithms with different step size parameters is designed. Thus, to a certain extent, the limitation of step parameters on the performance of the algorithm is alleviated. The simulation results in different environments show that the optimization strategy designed in this paper for blind source separation algorithm of momentum term can effectively eliminate the defects of the algorithm and obtain a small steady state error while ensuring fast convergence.
【作者单位】：烟台大学光电信息科学技术学院;
【基金】：国家自然科学基金(No.61005021,No.61201457) 山东省高等学校科技计划(No.J12LN27)
【分类号】：TN911.7

【参考文献】