摘要
EEG信号的非高斯特性导致了传统的EP信号提取算法的退化,为提高EP信号提取方法的韧性并实现少次提取,本研究利用BOREL谱测度的峰值确定欠定混合矩阵的基矢量,从而确定各个独立分量,并实现诱发电位的少次提取。仿真表明,利用基于BOREL测度的新方法分离前后的EP信号与EEG噪声的相关系数为0.9以上。这种方法是一种在分数低阶稳定分布噪声条件下具有良好韧性的诱发电位少次提取的新方法。
Non-Gaussian distribution of EEG leads to degeneration of traditional EPs extraction. In order to improve the robustness of the method of less trials extraction of EPs, a new method was proposed based on an estimate of the discrete spectral measure for identifying the independent components of an alpha-stable random vector for under-determined mixtures. Simulations results demonstrated that the method estimated the EEG and EP with correlation coefficient up to 0.9. The proposed method based on discrete BOREL spectral measure was robustness and could realized less trial extraction of evoked potentials.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2009年第2期177-182,共6页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金资助项目(60772037)
江西省卫生厅计划项目(20072048)
关键词
α-稳定分布
BOREL测度
诱发电位
独立分量分析
alpha-stable distributions
BOREL measure
evoked potentials
independent component analysis