摘要
文章建立了突发混合信号多通道盲分离模型,指出混合矩阵估计可以转化为混合信号幅度估计进行求解。突发信号通常属于恒模调制信号,其混合信号与未混合信号在幅度上存在明显差异。基于信号的恒模特性,提出了一种基于混合位置估计的多通道盲分离算法。该算法先通过混合信号的幅度最大最小值对混合位置进行估计,然后利用混合部位的最大最小值以及未混合部位的单个信号幅度值来进行幅度估计,提高了混合矩阵的估计精度。与传统盲分离算法相比,文章算法复杂度低,且消除了不确定性带来的不利影响。仿真表明,该算法所需数据量小,估计精度高,接近混合矩阵完美估计下的盲分离性能。
The model of multi-channel blind sourse separation （BSS） for burst mixing signals is es- tablished. Based on the model, the mixing matrix can be estimated by amphtude estimation. To con- stant module signals, the single signal and mixing signals vary a lot in amphtude, based on constant module character and hybrid position estimation, a new multi-channel blind sourse separation algo- rithm is proposed. Firstly, the hybrid position is estimated, using the max-min value of mixing part and amphtude of non-mixing part, the mixing matrix performance is better. Compared to traditional BSS algorithms, the proposed algorithm has no uncertainty problem but low complexity, The simula- tion shows that, the algorithm can be used in small data size situation, and its perfrmance approa- ches to mixing matrix perfect estimation.
出处
《信息工程大学学报》
2013年第3期320-324,340,共6页
Journal of Information Engineering University
基金
国家科技重大专项资助项目(2009ZX03003-005-01)
关键词
盲分离
恒模特性
幅度估计
混合位置估计
blind sourse separation
constant module
amphtude estimation
hybrid position estimation
