论文部分内容阅读
研究源信号数目未知与/或动态变化情况下的盲信号分离问题.首先证明若混合矩阵满列秩(观测信号的数目m不小于源信号的数目n),则互信息是盲信号分离的对比函数;在互信息的全局极小值点即分离点处,盲信号分离系统的输出除了零分量外,其他非零分量是希望提取的源信号.其次,利用混合矩阵的转置和m个观测信号向量构成的矩阵以概率1具有相同的零空间这一性质,只需少量观测样本就可以估计源信号的数目n,进而检测其动态变化情况.源信号数目未知且动态变化的盲信号分离计算机仿真验证了所提出理论和算法的有效性.
The problem of blind signal separation under unknown and / or dynamic change of source signal is studied.First, it is proved that mutual information is the comparison of blind signal separation if the mixed matrix is full column rank (the number m of observed signals is not less than the number n of source signals) Function, the non-zero component of the output of the blind signal separation system is the source signal to be extracted except for the global minimum of mutual information, ie the separation point.Secondly, using the hybrid matrix transpose and m observations The matrix composed of signal vectors is of the same null space as probability 1. With only a few observations, the number of source signals n can be estimated and the dynamic changes can be detected. A blind signal separation computer with unknown and dynamic source signals The simulation verifies the validity of the proposed theory and algorithm.