摘要
神经元锋电位(spike)是研究大脑信息编码的基础,因其宽带、小幅值等特点而极易受噪声干扰。针对spike信号的间歇性及非平稳性,采用经验模态分解(EMD,Empirical Mode Decomposition)的改进算法--整体平均经验模态分解(EEMD,Ensemble Empirical Mode Decomposition)对spike检测信号进行分解并结合小波阈值法进行去噪。EEMD能将信号中间歇性成分有效分离出来,解决了EMD产生的模态混叠问题。基于仿真和实测数据将其与EMD去噪方法及多元小波去噪法进行比较,结果表明:EEMD去噪方法不仅有效提高了spike检测信号的信噪比,而且降低了spike波形的畸变。在3种去噪方法中,EEMD去噪方法效果最为显著,对仿真信号的信噪比平均提高了4.177 2d B。为随后spike信号的分类和信息编码奠定了良好基础。
Spikes which are the basis of the research of brain information are sensitive to noise because they are broadband and small amplitude signal. Based on the fact that spikes are intermittent and nonstationary signals, EMD's improved algorithm EEMD was adopted to remove noise from neuronal spike signals with wavelet-threshold method. EEMD can solve EMD's model mixing by separating the intermittent composition in the signal effectively. Comparing with EMD with wavelet-threshold and Multivariate Wavelet, the result of simulation and real data shows that this method can not only improve SNR but also reduce spike waveform distortion. Among the three denoising methods, EEMD is the most effective by improving an average of 4.177 2 db in SNR. It is important for the detection and the next step analysis research of spike.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2015年第1期118-124,共7页
Journal of System Simulation
基金
国家自然科学基金(60971110)
河南省科技攻关计划项目(122102210102)
