基于多模态信息融合的心理负荷评估

为了相对稳定地识别人员是否处于心理负荷状态，设计并实施含能材料起爆作业心理负荷诱导试验。首先对心理负荷诱导情况进行判断，然后分别采集27名被试在静息状态和实施含能材料撞击起爆作业状态下的眼动、心率变异性(HRV)和脑电信号(EEG),通过正态性检验和假设检验获得心理负荷表征指标并进行统计功效分析，依据表征指标，采用支持向量机(SVM)和随机森林(RF)算法建立多模态信息融合的心理负荷评估模型，最后采用被试工作特征曲线(ROC曲线)分析各模态组合和分类器的心理负荷识别性能。研究结果表明：双模态(眼动+EEG)下SVM算法和3模态下RF算法评估性能和稳健性较高，多模态信息组合具有优异的识别效果。

Mental load assessment based on multi-modality information fusion

In order to relatively stably identify whether a person was in a state of mental load,the blasting operation of energetic materials was designed and implemented as the mental load induction test.Firstly,the induction situation of mental load was judged.Then,the eye movement,heart rate variability (HRV) and electroencephalo-gram (EEG) of 27 subjects were collected respectively in the resting state and the state of implementing the impact detonation operation of energetic materials.Thenormality test and hypothesis test were used to obtain the mental load characterization indexes,and the statistical efficacy analysis was conducted.According to the characterization indexes,the support vector machine (SVM) and random forest (RF) algorithms were used to establish the mental load assessment model with multi-modality information fusion.Finally,the receiver operating characteristic (ROC) curve was used to analyze the recognition performance ofmental load of each modal combination and classifier.The results showed that the SVM algorithm under dual-modality (eye movement + EEG) and the RF algorithm under three-modality had better assessmentperformance and robustness,and the combination of multi-modality information had an excellent recognition effect.