基于云模型的飞行员认知负荷分类方法研究

目的识别飞行员认知负荷状态,减少人为差错造成的事故。方法设计模拟飞行任务,诱导22名受试者产生高、低2种认知负荷状态。依据眼动存在的随机性和模糊性,建立基于凝视时间和瞳孔直径的二维云模型,根据该眼动指标的数学特征,构建了40条定性判定规则。根据规则生成器原理,构建了云模型定性推理生成器,对110组实验数据进行判断。结果受试者在高、低2种不同认知负荷下,凝视时间和瞳孔直径差异显著(P<0.05),采用定性推理生成器对认知负荷进行判定,平均准确率达到78.95%。在同样训练数据情况下,识别准确率高于K最近邻(KNN)算法和支持向量机(SVM)算法。结论定性推理生成器可有效识别认知负荷不同水平,且随受试者人数增加,识别率可进一步提高。

Cognitive Load Classification Method for Pilots Based on Cloud Model

Object To identify high cognitive workload of pilots and decrease accidents caused by human error.Methods Simulated flight mission was designed to induce high and low cognitive load status of 22subjects.Based on the randomness and fuzziness of eye movement,a two-dimensional cloud model based on fixation time and pupil diameter of subjects was established.According to the mathematical characteristics of these two eye movement indicators,forty qualitative determination rules were constructed.On the basis of the principle of rule generator,aqualitative model of cloud model was constructed,and 110sets of experimental data were judged.Results The fixation time and the diameter of pupil had an obvious difference(P<0.05).The cognitive load was determined by qualitative reasoning generator,and the average accuracy reached 78.95%.In the case of the same training data,the recognition accuracy was higher than the K nearest neighbor(KNN)algorithm and the support vector machine(SVM)algorithm.Conclusion The qualitative reasoning generator can be used for the detection of cognitive load,and the recognition rate can be further improved as the number of participants increases.