基于脑电功率谱-连续隐马尔科夫链的精神疲劳分级模型

Mental Fatigue Staging Model Based on Electroencephalogram Power Spectrum and Continuous Hidden Markov Model

提取多通道脑电(EEG)功率谱特征,训练连续高斯密度混合隐马尔科夫模型(CHMM),建立了基于功率谱-CHMM的精神疲劳分级模型.分级结果表明:EEG各节律功率谱及其比值是精神疲劳的敏感指标,CHMM对于不同的精神疲劳状态具有较高的分类精度,最高分类正确率达到97.5%;在训练样本相同的情况下,CHMM比反向传输人工神经网络具有更高的分类精度.

Multi-channel electroencephalogram（EEG） power spectrum is extracted for training continuous hidden Markov model （CHMM）, and a novel approach to classify the mental fatigue levels is proposed based on power spectrum-CHMM. The result shows that EEG power spectrum and the ratio of different rhythm serve as sensitive indices for mental fatigue, CHMM is effective for classifying metal fatigue levels with the highest classification accuracy of 97.5%. CHMM enables to classify more accurately comparing with back propagation neural network for the same training samples.