一种基于信息最大化的自适应变步长盲源分离算法

现有的盲源分离算法不适合于数据的实时处理,并且算法性能依赖于步长的选择.提出一种基于信息最大化的自适应变步长盲源分离算法,采用基于估计函数的变步长算法,降低了盲源分离算法性能对步长的依赖性,并且采用自适应处理形式,适合数据的实时处理.最后将其应用于声音信号的盲分离,在选择小的步长参数的情况下,原有算法和文中新算法都取得了良好的分离效果;在选择较大的步长参数的情况下,新算法优于传统算法.

Adaptive variable step-size blind source separation algorithm based on information maximization

The behavior of the classic algorithms for blind source separation(BSS)is not competent to real-time process and depends greatly on a fixed step size.It is difficult to choose a pretty step size.In this paper,the algorithm of blind source separation with a variable step size based on estimation function is proposed to improve the performance of the traditional algorithms.The step-size of the algorithm is adjusted to fit real-time process.Then,the two algorithms are used in blind source separation of sound signal,when choosing the step-size with little value,both two algorithms can work well;while in the situation of step-size with large value,the performance of the new algorithm proposed here is excellent than that of the old one.