论文部分内容阅读
利用信息论原理，推导出一个衡量输出分量独立性的目标函数，最小化该目标函数并利用信号的非平稳特性，得到一种可以进行非平稳信号的盲分离的训练算法；指出了在特定情况下，给出的两种网络结构形式是等价的，由此得到的训练算法避免了矩阵求逆．计算机仿真结果表明了该算法的有效性
Using the theory of information theory, an objective function is proposed to measure the independence of output components. The objective function is minimized and the non-stationary characteristics of the signal are utilized to obtain a training algorithm that can perform blind separation of non-stationary signals. It is pointed out that under certain conditions , The two network structures given are equivalent and the training algorithm thus obtained avoids matrix inversion. Computer simulation results show the effectiveness of the algorithm