Li Xin, Wang Huihui, Wang Yueru, Zhao Fangfang
Institute of Biomedical Engineering, Yanshan University, Qinhuangdao 066004, China.
Sheng Wu Yi Xue Gong Cheng Xue Za Zhi. 2011 Dec;28(6):1098-102.
According to the frequency overlapping of intrinsic mode function (IMF) based on the temporal and spatial filtering of empirical mode decomposition (EMD), which will lead to the question of useful signals and noises filtered together, we proposed a method that numbers of IMF is determined by energy estimate, temporal and spatial filtering combing wavelet threshold and EMD integrating wavelet local signal characteristics of time and scale domain. This method not only used multi-resolution wavelet transform features, but also combined the EMD and Hilbert decomposition of the adaptive spectral analysis of instantaneous frequency and significance of the relationship between energy, so as to solve the problem of useful signal being weakened. With MIT/BIH ECG database standard data subjects, experimental results showed it was an effective method of data processing for handling this type of physiological signals under strong noise.
基于经验模态分解（EMD）的时间和空间滤波的本征模函数（IMF）频率重叠会导致有用信号和噪声一起被滤除的问题，为此我们提出了一种方法，即通过能量估计、结合小波阈值的时间和空间滤波以及融合时间和尺度域小波局部信号特征的EMD来确定IMF的数量。该方法不仅利用了多分辨率小波变换特征，还结合了EMD和希尔伯特分解以进行瞬时频率的自适应谱分析以及能量关系的显著性分析，从而解决了有用信号被削弱的问题。以麻省理工学院/波士顿儿童医院心电图（MIT/BIH ECG）数据库标准数据对象进行实验，结果表明该方法是处理强噪声下这类生理信号的一种有效数据处理方法。
