一种针对欠定瞬时混合模型的盲辨识算法

Blind identification algorithm for the under-determined instantaneous hybrid model

文中研究了欠定瞬时混合模型下的盲辨识问题,提出了一种基于单源点检测的盲辨识算法。首先,文中介绍了欠定盲源分离的盲辨识模型;其次,利用观测信号时频系数之间的关系和大小来检测用于聚类的单源点;最后,采用势函数聚类的方法估计出混合矩阵。仿真结果表明,文中算法与其它相同混合模型的算法相比,具有更好的算法性能。

This paper researches the blind identification problem of the under-determined instantane-ous hybrid model and proposes a blind identification algorithm based on single source points detec-tion.First,this paper introduces the blind identification model of under-determined blind source separation.Then,single source points are detected through utilizing the relationship and values of the observed signals’time-frequency coefficients.Finally,the potential function clustering is used to estimate the mixing matrix.The simulation results prove that the proposed algorithm has better algorithm performance than other algorithms with the same hybrid model.