EEMD-ICA在功能性近红外光谱特征信号提取中的应用

发布时间：2019-02-18 10:35

【摘要】：近年来,功能性近红外光谱技术(fNIRS)广泛应用于神经影像学领域。为解决fNIRS特征信号提取中的信噪频谱混叠问题,依据近红外光谱脑功能成像信号非线性与非平稳特点,提出一种结合集合经验模态分解法和独立成分分析的多分辨率联合信号提取方法 EEMD-ICA。在脑功能成像仪器平台上采集多通道多波长脑功能成像近红外光密度信号,先对该信号进行集合经验模态分解将其按频率成分分解为多层本征模态函数,之后将独立成分分析应用于目标频率分量函数进行自适应去噪,最后将处理后的分量累加、重构获得近红外光谱脑功能成像的特征信号。将Valsalva氏实验测试数据作为研究对象进行滤噪处理,与经验模态分解法和集合经验模态分解法对fNIRS特征信号的提取效果对比。对实测数据的处理结果进行信噪比和误差参数分析,结果表明,该方法能够有效解决去噪过程中丢失原始信号有用信息及由于信噪频谱混叠不能完整去除噪声的问题,信号处理效果理想,对比另外两种信号提取方法更为优化。
[Abstract]:In recent years, functional near infrared spectroscopy (fNIRS) has been widely used in the field of neuroimaging. In order to solve the problem of signal noise spectrum aliasing in fNIRS feature signal extraction, according to the nonlinear and non-stationary characteristics of near infrared spectrum brain functional imaging signal, A Multiresolution Joint signal extraction method (EEMD-ICA.) based on set empirical Mode decomposition (EMD) and Independent component Analysis (ICA) is proposed. The signal of near infrared light density of multi-channel and multi-wavelength brain functional imaging was collected on the platform of brain functional imaging instrument. The signal was decomposed into multilayer intrinsic mode function according to the frequency component by means of set empirical mode decomposition. Then the independent component analysis (ICA) is applied to the target frequency component function for adaptive denoising. Finally, the processed components are accumulated and reconstructed to obtain the characteristic signals of NIR functional imaging. The experimental data of Valsalva's experiment were used as the research object for noise filtering, and compared with the empirical mode decomposition (EMD) and set empirical mode decomposition (EMD) to extract the fNIRS feature signal. The signal to noise ratio (SNR) and error parameters of the measured data are analyzed. The results show that the method can effectively solve the problem of missing the useful information of the original signal during the process of de-noising and the problem that the signal and noise spectrum can not be completely removed because of the aliasing of the signal and noise spectrum. The effect of signal processing is ideal, compared with the other two signal extraction methods are more optimized.
【作者单位】：吉林大学仪器科学与电气工程学院;吉林大学第一医院;
【基金】：国家重大科学仪器设备开发专项子课题(120111222867002) 国家自然科学基金项目(21207047) 高等学校博士学科点专项科研基金项目(20120061110092) 吉林省科技发展计划重点项目(20120328)资助
【分类号】：R741.044;TN911.7;O433.4

【参考文献】