【摘要】 清醒期 (W )、快速眼动期 (REM)和睡眠二期 (S2 )在睡眠总时间中占据很大比例 ,而且三者从脑电 (EEG)上较难区分。用隐马尔可夫模型 (HMM)从单导睡眠脑电中区分W期、REM期和S2期。对受心电干扰明显的脑电信号进行独立分量分析 (ICA) ,去除干扰 ;建立最佳阶数AR模型 ,进行谱分析 ,提取EEG平均频率 ,和EEG幅度均值、标准差一起作为观察值 ;分别建立W期、REM期和S2期的连续密度隐马尔可夫模型 (CD -HMM )。经过测试 ,W期、REM期和S2期的正确识别率分别为 92 % ,10 0 %和 94%。表明隐马尔可夫模型 (HMM)在睡眠分期中有很好的应用前景更多还原

【Abstract】 Sleep stages-Wake,REM and S2 occupy large ratio in all sleep time.It is difficult to distinguish them only by EEG. In this paper,HMM is applied to distinguish Wake,REM and S2 by one channel EEG. EEG was preprocessed using Independent Component Analysis(ICA) to eliminate obvious ECG interfere,then took spectral analysis by optimal order AR model.EEG mean frequency(MF),amplitude average value and standard deviatio were calculated as observation vector.A continuous density HMM was established for Wake,REM and S2 respectively.Evaluation shows that the recognition rate of Wake,REM and S2 are 92%,100% and 94% respectively.It means HMM is very useful in sleep staging.更多还原