星载AIS的复值FastICA算法改进

针对星载船舶自动识别系统(AIS)的含噪复值信号盲分离算法分离效果不佳的问题,提出了改进的复值快速独立分量分析算法（FastICA）。该改进算法针对混合信号数目大于源信号数目的超定情况,对含噪混合信号的协方差矩阵进行特征值分解,利用其噪声对应的几个较小特征值估计噪声方差,修正白化矩阵,再应用Huber M估计函数优化该算法的目标函数。实验结果表明,运用该算法信号均方误差(SMSE)变小,信干比(SIR)变大,提高了信号的分离性能；同时,优化后的目标函数使算法具有良好的稳健性。

Improvement of complex fast independent component analysis algorithm for satellite-based AIS

For the problem that the performance of blind separation for noised satellite-based automatic identification system(AIS) signal is not satisfying, an improved complex fast independent component analysis(FastICA)algorithm is presented. The proposed algorithm obtains the covariance matrix by the mixed signals under overdetermined case in which the number of mixed signal is greater than that of source signal,and then the covariance matrix is decomposed to get the eigenvalues,the small ones to which noise corresponds are used to estimate the approximate noise variance and improve the whitening matrix,and Huber M-estimation function is applied to optimize objective function of the algorithm. The experimental results show that the proposed algorithm has lower signal mean square error(SMSE) and larger signal to interference ratio(SIR),so it can improve the separation performance. At the same time,the optimized objective function makes the algorithm robust.