基于样本选择的跨被试脑力负荷识别

Cross-subject Mental Workload Recognition Based on Instance Selection

人体的脑力负荷状态与人机操作工作时的工作效率、人力资源分配以及事故的发生等息息相关,因此研究操作人员的脑力负荷状态具有重要意义。为了解决现有脑力负荷识别方法由于训练集中样本数量过少导致分类效果较差的问题,提出了一种基于样本选择的跨被试脑力负荷识别方法。首先,将其他被试的脑电数据作为训练集,参考目标被试的少量历史数据对训练集中的特征数据进行样本选择,实现减少样本数量的同时减少训练集和测试集之间的域差异;其次,再通过主成分分析对样本选择后的自适应训练集和目标被试测试集特征进行特征降维;最后,用自适应训练集主成分建立支持向量机分类模型识别测试集样本的脑力负荷状态。结果表明,该方法可以在提高分类效率的同时提高分类精度,实现快速、准确的脑力负荷状态识别。

The degree of mental workload is closely related to the work efficiency and the allocation of human resources and the occurrence of accidents of man-machine operation.Therefore,it is of great significance to study the state of operators'mental workload.In order to solve the problem that the classification effect of the existing mental workload recognition methods is poor due to the small number of samples in the training set,a cross-subject mental workload classification method based on instance selection for visual and operational task was proposed.Firstly,referring to a small amount of historical data of the target subject,an adaptive source domain training set was obtained by instance selection of the training set data of the electroencephalogram(EEG)data of other subjects,thereby the number of samples and the domain difference between the training sets and test sets was reduced.Secondly,principal component analysis was used to reduce the feature dimension of the adaptive training set and target test set.Finally,the adaptive training set principal component was used to establish a support vector machine classification model to identify the mental workload state of the test set samples.The results show that this method can improve classification efficiency while improving classification accuracy,and achieve fast and accurate mental workload state recognition.