【摘要】 许多盲源分离算法对超弱信号收敛速度变慢甚至失效。文章提出一种算法 ,用FASTICA算法所得的分离矩阵作为一种基于输出去相关自适应盲分离算法的初始值 ,算法的收敛速度得到明显改善。算法的学习速率通过一种自适应算法选取 ,从而降低了算法收敛速度对学习步长的依赖。仿真结果表明 ,文中提出的算法对超弱信号特别有效 ,算法收敛速度大大增强。该算法可以用于分离强干扰下的弱水声信号更多还原

【Abstract】 Either most of algorithms for blind source separation fail with the super-weak signal, or convergence speed of the algorithm is slow. This paper presents an adaptive algorithm as follows. Using the separate matrix yielding from the FASTICA algorithm as original separate matrix may improve the converge speed of the adaptive algorithm which based on decorrelation of output signals. Learning rate were chosen by an adaptive algorithm such that the convergence speed wasn’t sensitive to learning step. Computer simulation demonstrates the algorithm is good especially for super-weak signals. The convergence speed of present algorithm is higher in contrast to other algorithms. Present algorithm can be used to separate super weak underwater acoustics signals in the stronger jammer.更多还原