PAC耦合的脑电信号中疲劳特征过滤提纯仿真
Simulation of Fatigue Feature Filtering and Purification in PAC Coupled EEG Signals
张晨 1杨硕1
作者信息
- 1. 河北工业大学生命科学与健康工程学院,天津 300130;河北省生物电磁与神经工程重点实验室,天津 300130;天津市生物电工与智能健康重点实验室,天津 300130
- 折叠
摘要
脑电信号具有高维度和复杂性,如果筛选出的特征不合理,会导致分析结果存在较大的误差.针对这一问题,研究一种脑电信号中疲劳相关特征过滤提纯方法.针对采集到的脑电信号,利用ICA方法去除其中的伪影,降低非脑电活动所引起的信号的干扰,并通过计算信息增益实现脑电信号中疲劳相关特征过滤.针对过滤出来的功率谱密度以及PAC耦合值特征,通过计算熵值进一步筛选出不同波段的功率谱特征,完成特征提纯.测试结果表明:所研究方法应用提纯得到的功率谱特征+PAC特征输入下,准确率相对较高,且反应时间相对较低,由此说明所研究方法过滤提纯得到的功率谱特征和PAC特征较为合理.
Abstract
EEG signal has high dimensionality and complexity,and unreasonable features may lead to significant errors in the analysis results.To address this issue,a method for filtering and purifying relevant features of fatigue in EEG signals was researched.For the collected EEG signals,ICA method was used to remove their artifacts,thus re-ducing interference caused by non-EEG activities.Meanwhile,fatigue features in EEG signals were filtered by calcu-lating information gain.According to the filtered power spectral density and PAC coupling value,the power spectral characteristics in different bands were further screened by calculating entropy.Finally,the feature purification was completed.Following conclusions can be drawn from the test results:based on the proposed method,the accuracy of purification under the power spectrum features and PAC features input is relatively high,and the reaction time is low-er,indicating that the power spectrum features and PAC features obtained by the method are more reasonable.
关键词
脑电信号/疲劳相关特征/特征过滤/特征提纯/功率谱密度Key words
EEG signal/Related characteristics of fatigue/Feature filtering/Feature purification/ICA method/Power spectral density引用本文复制引用
出版年
2024
NETL
NSTL
万方数据