基于脑电的水平运动立体影像视疲劳评估

Visual Fatigue Assessment of Stereoscopic Images in Lateral Motion Based on EEG

针对长时间观看水平运动立体视频所引发的视觉疲劳这一现象,采用单一运动形式的水平立体视频刺激为实验素材,通过主观实验对比分析被试观看同样的3D素材与2D素材后所引起疲劳感受的差异性.分别采集被试疲劳前后的EEG(electroencephalography)信号与ERP(event related potentials)信号,提取疲劳前后30个电极EEG信号的θ、α、β3个波段小波系数的均值及其比值、、θ/(α+β)、(α+θ)/β、θ/β、(α+θ)/(α+β)作为时频域特征,提取样本熵作为脑电信号的非线性特征.将时频域特征与非线性特征作为联合特征进行立体视觉疲劳评估,避免了单一特征的局限性,多角度分析了视觉疲劳前后特征参数变化.实验结果表明,3D与2D视频均能够使被试在一定程度上产生视觉疲劳现象,且3D显示比传统2D显示对人的影响更大.对比分析ERP信号发现疲劳状态下的N200和P300成分幅值明显下降,而潜伏期略有增加.通过可分性指数FCSI(Fisher’s class separability index)算法、支持向量机SVM(support vector machines)、基于互信息的特征选择算法等筛选出评估水平运动视觉疲劳的最佳特征参数为CP3电极上的α/θ.同时,结合脑区分析发现,与水平运动立体视频关系最为密切的两个脑区可能为顶区与前额区.

For stereoscopic visual fatigue caused by watching stereoscopic video clips in lateral motion for a long time,horizontal stereo video stimulation of a single motion form was used as experimental material.The differences in fatigue feeling caused by watching the same 3 D and 2 D materials were compared and analyzed using subjective questionnaires.The electroencephalography and event related potential(ERP)signals before and after fatigue were seperated.The features,including the mean of the wavelet coefficients of θ,α,and β and the ratios of α/β,α/θ,θ/(α+β),(α+θ)/β,θ/β,and(α+θ)/(α+β) from each electrode were extracted as time–frequency domain features and sample entropy as a nonlinear feature.The time–frequency domain feature and nonlinear feature were used as joint features for stereoscopic visual fatigue assessment,which avoids the limitation of a single feature.The characteristic parameters before and after visual fatigue from multiple angles were analyzed.The experimental results show that both 3 D and 2 D videos can cause the visual fatigue phenomenon to a certain extent,and the 3 D display has more influence on people than the traditional 2 D display.The N200 and P300 components of the ERP signal before and after fatigue were compared and analyzed.It is observed that the amplitudes are significantly decreased,whereas the latencies are slightly increased.Using Fisher’s class separability index algorithm,support vector machine,and feature selection algorithm based on mutual information,we determined that the index of α/θ in the CP3 electrode is appropriate to evaluate the visual fatigue caused by watching the stereoscopic video in lateral motion.At the same time,combined with brain area analysis,it is found that the two brain regions most closely related to the lateral motion stereoscopic video may be the top and forehead regions.