基于支持向量机的欠定盲分离

SVM Based Underdetermined Blind Source Separation

该文提出了信号稀疏性的新度量方式,在估算出有效源信号的个数后,提取源信号到达方向角度的特征作为训练样本,利用支持向量机理论构造分类超平面,从而实现对观测信号的最优分类。采用加权系数法获得每一类信号的聚类中心,其中对系数权重的学习是自适应的,同时避免了K-均值聚类等方法对初值的敏感性。此外,针对大规模样本点,该文还提供了在线算法。仿真效果说明了此方法的稳定性和鲁棒性。

A new sparse measure of signals is proposed in this paper. After the number of efficient sources is estimated, the observations are classified using Support Vector Machine （SVM） trained through samples which are constructed by the direction angles of sources. Each clustering center is obtained based on the sum of samples belong to the same class with different weights which are adjusted adaptively. It gets out of the trap of the initial values which interfere k-mean clustering seriously. Furthermore, the online algorithm is proposed for large scale samples. Simulations show the stability and robustness of the methods.