论文部分内容阅读
在盲信号分离中，常用的方法假设源信号的密度函数已知或由某一类带参数的函数来逼近，一旦假设的密度函数和真实的密度函数不一致或源信号不是同一分布，所用的方法就不能正确地分离出源信号。针对这一问题，本文用高斯混合模型来逼近任意分布的源信号的密度函数，并提出了一种迭代的期望最大化算法。计算机仿真表明，该算法能够有效地分离出真实的源信号。
In blind signal separation, the commonly used method assumes that the density function of the source signal is known or approximated by a type of parameterized function. Once the assumed density function is not consistent with the true density function or the source signals are not the same distribution, the method used The source signal can not be correctly separated. In order to solve this problem, this paper uses Gaussian mixture model to approximate the density function of arbitrary distributed source signals and proposes an iterative expectation maximization algorithm. Computer simulation shows that the algorithm can effectively separate the real source signal.