超定条件下的盲信号提取算法

Blind signal extraction algorithm in overdetermined mixtures

基于信号峭度理论,提出一种超定条件下的盲信号提取算法。该算法将混合矩阵辨识转化为一系列Givens矩阵辨识,从观察信号中一次提取出一个源信号。对于超定盲信号分离问题,待未知所有独立分量分离出后,余下分量可以看作是一个或多个独立分量的拷贝,是冗余信号。在算法运行结束后,所有源信号分离出,实现超定盲信号分离。该算法计算简单,收敛性好。计算机仿真试验验证了算法的有效性。

Based on the signal kurtosis, a blind signal extraction algorithm was proposed in over-determined mixtures. Converting the mixture matrix into a series of Givens matrix identifications, the algorithm can extract signals one by one from the mixtures. When the all unknown independent components were separated, the other components were redundant signals that can be regarded as copies of one or some independent components. All source signals were separated while the algorithm was accomplished. The algorithm is easy to implement with fast convergent speed. Computer simulation confirms its validity.