Computer Engineering and Applications ›› 2016, Vol. 52 ›› Issue (14): 130-135.

Previous Articles     Next Articles

Inhibition of edge effect of adaptive single channel blind source separation

WU Longhua1, ZHU Jiagang1, LU Xiao2   

  1. 1.School of IOT Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
    2.Co-Laboratory in Hillsun Ltd., Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2016-07-15 Published:2016-07-18

抑制边缘效应的自适应单通道盲源分离

吴龙华1,朱嘉钢1,陆  晓2   

  1. 1.江南大学 物联网工程学院,江苏 无锡 214122
    2.江南大学 晓山股份联合实验室,江苏 无锡 214122

Abstract: Single Input Multiple Out Blind Source Separation(SIMO_BSS) is a special kind of underdetermined blind source separation. To address this problem of single channel, it usually takes the Ensemble Empirical Mode Decomposition and Independent Component Analysis combination algorithms(EEMD_ICA). However, on the basis of the EEMD blind source separation algorithm, it produces the edge effect to reduce the signal separation accuracy. In this paper, it presents a method to suppress the edge effect through increasing prediction extreme value points, this method on the time and space complexity is significantly superior to the method based on the cycle continuation of source signals, and it is suitable for the long sequence signal. Under different SNR, through ECG mixed signal simulation, the separated performance of this method is compared with EEMD-ICA and EEMD-PCA-ICA algorithms, and experimental results show that this method outperforms the two latter algorithms with higher correlation coefficient. Finally, the practical application of this algorithm to the piezoelectric periodic signal, results show that this algorithm has obvious separation effect.

Key words: single channel blind source separation, edge effect, ensemble empirical mode decomposition, principal component analysis, independent component analysis

摘要: 单入多出的盲源分离SIMO_BSS是一种特殊的欠定盲源分离。针对单信道问题,通常采用总体经验模态分解和独立成分分析联合使用EEMD_ICA算法。然而,以EEMD为基础的盲源分离算法,会产生边缘效应降低信号分离准确率。因此,提出了一种在端点处增加预测极值点的方法来抑制边缘效应,在时间、空间复杂度上要明显优于基于周期延拓源信号的方法,而且适用于长序列信号的分离。在不同的信噪比SNR下,通过心电ECG混合信号仿真,该方法比EEMD_ICA方法,以及EEMD_PCA_ICA方法分离出的信号相似度高。最后将该算法实际应用到周期压电信号中,结果表明该方法具有明显的去噪分离效果。

关键词: 单通道盲源分离, 边缘效应, 总体经验模式分解, 主成分分析, 独立成分分析