基于自然梯度算法的盲信源分离研究

盲信源分离试图从给定的一组混合观察数据中恢复未知的独立信源。文中介绍盲信源分离的一种非常重要的算法——自然梯度算法。对通信信号和自然语音信号采用不同的活动函数进行了盲信源分离的计算机模拟实验，结果显示该算法能够分别有效地分离这两类随机混合的信号。

Research on Blind Source Separation Based on Natural Gradient Algorithm

Blind source separation attempts to recover unknown independent sources from a given set of observed mixtures. The natural gradient algorithm is introduced in this paper, and it is a very important approach for blind source separation. We have examined the algorithm with communication signals and natural speech signals by different activation functions respectively. Simulation results demonstrate the algorithm can effectively separate the two kinds of random mixing signals.