针对之前盲源抽取算法存在鞍点的问题,提出一种新的基于二阶统计量的盲源抽取算法。通过利用自回归模型对抽取信号向量进行估计,并利用估计值与抽取向量之差提出一种新的代价函数,证明了代价函数的有效性。通过利用最速下降法对抽取向量以及FIR滤波器权值向量的计算,求解出抽取向量最优值。最后通过仿真证明算法相对之前两种算法有更高的可靠性,且在低信噪比的环境下,算法抽取效果依然良好且保持很高的抽取正确率。

In view of saddle points in blind source extraction algorithm before, an improved blind source extraction algorithm was proposed. Firstly, target signal vector was estimated by using autoregressive model, and a new cost function was proposed by using difference between the estimated vector and the extracted signal vector, effectiveness of new cost function was proved. The optimal solution of the extracted vector was computed by using the steepest descent method to extract vector and weight vector of FIR filter. At last, computer simulation proved that the improved algorithm shows higher reliability than the other two algorithms, and with better extraction effect and keeps higher accuracy in low SNR environment.