【摘要】 提出了一类新的实时线性混叠信号盲分离算法,该算法基于混叠矩阵的QR分解思想,结合均匀分布的源信号相互独立时其Scatter图具有的特殊形状以及与坐标轴平行的几何性质,导出了一类盲源分离的几何算法.本文的方法相对Taro和Hyvarinen而言,大大简化了其分离过程,从而缩短了分离时间.仿真结果表明,对同样两幅混叠图像的分离,在效果相当的情况下,本文算法的分离时间比Hyvarinen的分离时间缩短了约2.5倍.更多还原

【Abstract】 Geometric algorithms for blind signal separation(BSS) have recently received some attention due to their pictorial description and their relative ease of implementation.We present a new blind separation algorithm of instantaneous mixing (model,)which is based on QR decomposition of mixing matrix.And the scatter plot of source signals show special geometrical shape when source signals are statistically independent and uniformly distributed.Comparing with Taro or Hyvarinen algorithms,this algorithm is very simple and fast.Simulation result shows that the separation time of our algorithm is two and half times shorter than that of Taro and Hyvarinen under the same separation performance and the same source signals.更多还原