论文部分内容阅读
首先定义了描述信号分离状态的信号相依性测度,并利用此测度将传统算法中的学习速率参数推广至二维矩阵,从而提出了一种基于分阶段学习的盲信号分离算法,即整个信号分离过程被分成三个阶段进行:初始阶段、捕捉阶段和跟踪阶段,每个阶段的学习速率由信号的分离程度自适应选取.理论分析表明,该算法满足等变化性和分离矩阵的非奇异性条件.仿真结果证实,新算法具有比使用固定和其他自适应学习速率的算法更快的收敛速度、更好的稳态性能和更高的数值稳定性.
Firstly, the signal dependence measure, which describes the state of signal separation, is defined and used to extend the learning rate parameter in the traditional algorithm to a two-dimensional matrix. A blind signal separation algorithm based on phased learning is proposed, that is, the whole signal separation The process is divided into three stages: the initial stage, the capture stage and the tracking stage, the learning rate of each stage is adaptively selected by the degree of signal separation.Theoretical analysis shows that the algorithm satisfies the variability and the singularity of the separation matrix The simulation results show that the new algorithm has faster convergence speed, better steady-state performance and higher numerical stability than the algorithm using fixed and other adaptive learning rates.