【摘要】 单通道盲源分离(SCBSS)技术是在未知任何先验信息的条件下,仅由一路接收信号估计出多路源信号的信号处理方法,目前的SCBSS算法仍没办法完全精确地分离出所有源信号。为了提高部分源信号的分离精度,提出一种基于双树复小波变换(DTCWT)的单通道盲源分离算法。算法先用DTCWT对混合信号进行分解,再由PCA从分解信号中筛选出数目比源信号数目少一个的信号分量,这些分量与混合信号一起构成虚拟的多通道信号,最后用Fast-ICA估计出各个源信号。上述方法极大程度地减少了传统小波分解中的频率混叠问题。实验证明,和基于传统小波分解的单通道盲源分离算法相比,上述算法的分离性能得到了明显的改善。更多还原

【Abstract】 Single-channel blind source separation(SCBSS) technology is a signal processing method in which only one received signal is used to estimate the source signals without any prior information, and the current SCBSS algorithms are still unable to separate all the source signals completely and accurately. In order to improve the separation accuracy of partial source signals, this paper proposes a single-channel blind source separation algorithm based on dual-tree complex wavelet transform(DTCWT). The mixed signal was decomposed by DTCWT, and then PCA was used to select the signal components with the number one less than that of source signals. These components were combined with the mixed signal to form virtual multi-channel signals. Finally, each source signal was estimated by Fast-ICA. This method greatly reduces the frequency aliasing problem in traditional wavelet decomposition. Experimental results show that the performance of the proposed algorithm is obviously improved compared with the single-channel blind source separation algorithm based on traditional wavelet decomposition.更多还原