摘要
根据母亲心电和胎儿心电在时域上具有的局部稀疏特性,提出了一种新的基于抑制源信号Wigner-Ville时频分布交叉项的胎心电信号半盲分离方法。该方法首先搜索腹部观测信号中母亲心电信号和胎儿心电信号局部稀疏的区域,再将该区域观测信号变换到Winger-Ville时频域,并在其中搜索源信号的子项和交叉项,以构建盲分离的目标函数提取胎儿心电信号。该方法将被认为无用的时频交叉项引入盲分离算法实现了胎儿心电信号的分离。模拟信号的实验证明,该算法与传统利用伪Wigner-Ville时频分布去除Wigner-Ville交叉项的盲分离方法性能非常接近。临床实验表明,该方法与FastICA方法和传统去除Wigner-Ville交叉项的盲分离方法获得的提取信号相似系数分别为0.953 3和0.970 8。
According to the local sparseness of maternal ECG（electrocardiogram） and fetal ECG signals,a new semi-blind source separation method is proposed to extract fetal ECG from the mixed abdominal observations.The first step of the method is to search the local sparse segment in the abdominal observations.Then,the method converts the sparse segment into the Wigner-Ville time-frequency domain,and obtains the auto-terms of the source signals and the cross-terms between them,and constructs the objective function to extract fetal electrocardiogram.The method introduces the cross-terms of spatial Wigner-Ville distributions,which are thought to be useless and must be removed by kernel functions before further processing,to the contrast function for fetal ECG extractions.Thus,the proposed method is different from traditional blind source separation method in Wigner-Ville time-frequency domain.Simulation experiment results depict that the performance of the algorithm is much close to that of traditional methods.The similarity coefficient between the results obtained using the proposed method and FastICA method is 0.953 3,and it is up to 0.970 8 between the results obtained using the proposed method and traditional BSS method in Wigner-Ville time-frequency domain.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2012年第4期781-787,共7页
Chinese Journal of Scientific Instrument
