一种基于极大似然估计的信号盲抽取算法

An Algorithm of Blind Source Extraction Based on Maximum Likelihood Estimation

摘要:

针对盲信号抽取问题,根据极大似然估计原理,给出一种基于极大似然估计的分步盲抽取算法。根据信号不同的延时特性确立抽取信号,完成第一步抽取。由于初步抽取信号存在噪声污染,根据极大似然估计,估计抽取信号的概率密度函数,确立优化代价函数,利用自然梯度方法进行优化,确立最终抽取向量迭代方式,完成对抽取信号优化处理。通过仿真证明算法具有良好的收敛性和抗噪性,在 SNR>17dB时,抽取信号与源信号的相似度达 到 95%。

Abstract:

Aim at the blind source extraction(BSE)question, an algorithm was proposed based on maximum likelihood es. timation(MLE). First, the extraction signal was ensured based on different time-lapse character, the first floor extractionwas completed. But, because of the jamming of noise in signal, the probability density function was estimated based onMLE, then the cost function was ensured, and the cost function was optimized by use the natural gradient method. At last,the extraction vector was ensured, and the process of optimize extraction signal was completed. The experiment validatedthe algorithm had better convergence character and better performance in strong noise environment. The similarity degreearrive at 95% when SNR>17dB.