稀疏分量分析的RANSAC算法

RANSAC algorithm of sparse component analysis

为了提高盲源分离的分离性能,提出了基于随机抽样一致(RANSAC)的稀疏分量分析方法。首先,如果时域源信号不稀疏,则先对混合信号按要求进行短时傅里叶变换、小波变换或非下采样轮廓波变换等线性变换,使其在相应的变换域稀疏化;其次,利用RANSAC算法以序贯的方式估计出混合矩阵的列向量;最后,利用最短路径法等源估计算法估计出源信号,实现盲源分离。理论分析和实验均表明,此算法不仅可以有效地估计出混合矩阵,而且对稀疏度要求适当降低的信号也有较好的分离效果。

In order to improve the performance of blind source separation,this paper proposed a new method of sparse component analysis （SCA） based on random sample consensus （RANSAC）.Firstly,if the source signals were not sparse in the time domain,then it decomposed the mixed signals by linear transform such as short time fourier transform（STFT）,wavelet transform（WT） or nonsubsampled contourlet transform （NSCT）.It made the signals sparse in the transform domain.Secondly,it computed the column vectors of mixed matrix successively by RANSAC algorithm.Finally it estimated source signals by classic methods such as shortest path method.Theory analysis and simulation experiments all show that the new method can not only estimate the mixed matrix effectively,but also achieve superior performance in the signals which are not sparse strictly.