基于信源信息熵最小的单通道盲源数估计算法-《郑州大学学报(工学版)》

文章信息/Info

Title:: Single Channel Blind Source Number Estimation Algorithm Based on Source Information Entropy Minimization

Author(s):: MAO Ling; ZHAO Lianwen; MENG Hua; LI Yukai; School of Mathematics, Southwest Jiaotong University, Sichuan Chengdu 610031

Keywords:: single channel blind source separation; source number estimation; information entropy; Gaussian mixture model; Markov chain Monte Carlo algorithm

摘要:: 源数估计问题是盲源分离(BSS)中的一个关键问题,因为源数会直接影响盲源分离的效果。针对此问题提出了一种将信息熵作为统计评价指标的单通道盲源数估计算法,使用信息熵来度量源信号的信息量大小从而确定源数。为了计算估计源信号的信息熵,首先,使用高斯混合模型( GMM) 来拟合其分布;其次,基于马尔可夫链蒙特卡罗(MCMC)算法,采样得到服从目标分布的样本,并进行熵的计算;最后,通过最小化估计源信号平均信息熵得到盲源个数。一系列基于仿真数据和真实通信数据的实验表明,所提算法具有较强的鲁棒性,且能以 94% 的准确率估计出源数,从而验证了算法的有效性。

Abstract:: The problem of source number estimation was a key issue in blind source separation (BSS) , because the number of sources directly affected the effect of BSS. To solve this problem, this paper proposed a single-channel blind source number estimation algorithm that took the information entropy as the statistical evaluation index, and used the information entropy to measure the information quantity of the source signal to determine the source number. To calculate the information entropy of the estimated source signals, firstly, the Gaussian mixture model (GMM) was used to fit their distributions; Secondly, samples obeying the target distribution were sampled and the entropy is calculated based on the Markov chain Monte Carlo (MCMC) algorithm; Finally, the source number was obtained by minimizing the average information entropy of the source signal. A series of experiments based on simulation data and real communication data show that the proposed algorithm has strong robustness and can estimate the number of sources with 94% accuracy, thus verifying the effectiveness of the algorithm.