基于线性聚类的稀疏成分分析及其在盲源分离中的应用

SPARSE COMPONENT ANALYSIS BASED ON LINEAR CLUSTERING AND APPLICATION TO BLIND SOURCE SEPARATION

提出一种基于线性聚类的稀疏成分分析法,给出相关理论证明和实现算法.该方法充分利用稀疏源信号的线性混合信号沿混合矩阵列向量方向线性聚类的特性进行盲源分离.实验结果表明,在独立成分分析失效的情况下(源信号相关或高斯分布)仍然能够有效地分离出潜在的稀疏源.对分离出的信号及源信号进行相关系数分析,分离出的信号与源信号完全线性相关.基于线性聚类的稀疏成分分析法能准确地重构稀疏源信号.

Sparse component analysis based on linear clustering was used for blind source separation, with related proof and algorithm given. The characteristic that linear mix of sparse source signals clusters along vectors of mixed matrix was made full use of. The experimental results indicate that this method can effectively extract source signals even when source signals are dependent or Gaussian distributed. According to correlation coefficient analysis, separation results and sources are linear correlated, indicating that linear clustering method can accurately reconstruct sparse source signals.