基于混合刺激诱发放松状态的方法研究

Study on the Method of Inducing Relaxation State Based on Mixed Stimulus

论文研究在视觉、听觉、混合刺激下,采集脑电信号,刺激诱发实验者进入高度放松状态的方法。文中使用LSTM,SVM,Random Forest等方法对刺激前后的脑电进行分类,通过学习得到分类器,并对调节阶段的数据进行切割找出不同刺激方式的最佳时间。通过该实验验证,1)采用滑动时间窗口方法,分别提取均值、方差、极值、频域、变化系数、波动指数等时域特征和功率熵、重心频率等频域特征取得较好效果;2)使用时间序列模型(LSTM)对脑电分类的准确率可以达到92%;3)根据有效刺激时间,混合刺激时间比听觉刺激缩短了30.6%,比视觉刺激缩短了7.7%;4)视觉刺激影响脑电波的Theta和Alpha频段,听觉刺激影响脑电波Delta频段。

In this paper,through the visual,auditory,mixed stimulation,the experimenter is adjusted into a highly relaxed state.Under these three kinds of stimuli,EEG signals are collected.In order to classify the data after and before stimulation by SVM,random forest,GBDT,LSTM methods,classifiers are used to cut the data of the adjustment phase to find out the effective time of different stimulation methods.The experiment verifies that 1)sliding time window is used to extract the mean,variance,extreme value,fluctuation index on time domain and power entropy,gravity frequency on frequency domain,have achieved good results.2)The effect of time series model(LSTM)on EEG classification is much better,and the accuracy rate can reach 92%.3)According to effective stimulation time,can get mixed stimulation effect better than visual stimulation,and visual stimulation effect better than auditory secondary thorn.4)Visual stimuli affects the Theta and Alpha bands of brain waves,and auditory stimuli affects the Delta bands of brain waves.