一种新的盲源分离拟开关算法

A new quasi switching algorithm for blind source separation

针对电子战中各种信号混叠严重难以分离的现象,在盲源分离开关算法基础上提出一种新的盲信源分离拟开关算法.该算法引入单位对称加权滑动向量来加权每次迭代所得的分离信号作为源信号,用峭度取代原算法的峭度符号位作为判断函数来自适应选择加权相应激活函数,以此优化学习算法,结合信号分选的具体应用,给出了迭代结束的评判方法.计算机仿真实验表明,在强噪声背景影响下,该算法能够更加有效地分离空间未知多源线性混叠信号,且在分离效果、稳定性、处理速度和抗噪性能上都比原算法有较大改进.

A new quasi switching algorithm for blind source separation （BSS） was developed that resolves difficulties in the separation of all kinds of seriously mixed signals in an electronic warfare （EW） environment. The algorithm first introduces a unit symmetrical sliding weighted vector to weight estimated signals after iteration as source signals. It uses signal kurtosis to replace a symbol bit for kurtosis as a judging function to adaptively choose and weight the corresponding activation function. This gives a new judging method which was used to judge the finishing time of iteration in signal sorting. Computer simulation results showed that, under the influence of strong background noise, the novel algorithm can be more effective in unknown multi-source separation of linear mixed signals than the original algorithm. In addition, separation efficiency, stability, process capacity and anti-noise performance were greatly improved relative to the original algorithm.