基于BDPCA的驾驶员脑疲劳等级划分研究  

Analysis of Driver Brain Fatigue Level Based on BDPCA

作　　者：吕超[1] 闫超[1] 徐亚茹 年锦涛 LV Chao;YAN Chao;XU Ya-ru;NIAN Jin-tao(Changchun University of Science and Technology,Changchun Jilin130021,China)

机构地区：[1]长春理工大学,吉林长春130021

出　　处：《计算机仿真》2022年第11期208-214,共7页Computer Simulation

基　　金：吉林省科技发展计划项目(20200401096GX)。

摘　　要：当人体产生疲劳状态时,大脑释放的脑电信号也会发生相应的变化。在以往对脑疲劳状态的研究中,研究者多从清醒与疲劳两种状态进行分析,忽略了对不同的疲劳状态程度的研究,且对不同疲劳状态划分的定义并不客观。针对脑疲劳状态等级划分研究不充分的问题,提出了一种基于非监督学习的聚类算法对疲劳状态等级进行客观性的划分。通过小波包分解提取脑电信号的节律能量和非线性特征作为特征向量,使用共同邻域参数(CNN)改进的DPCA聚类算法对提取到的特征向量进行分析训练。同时,使用贝叶斯准则(BIC)对类簇个数进行辅助判定。实验结果证明,改进后的BDPCA算法准确率可以达到85%以上,能够对脑电信号中表征的不同疲劳状态等级进行准确划分,实现了脑疲劳状态等级的客观性定义。When the human body is in a state of fatigue,the EEG signals released by the brain will also change accordingly.In the previous studies on brain fatigue,researchers mostly analyzed from the two states of wakefulness and fatigue,ignoring the study on different levels of fatigue,and the definition of different fatigue states is not objective.To solve the problem of insufficient research on the classification of brain fatigue status,a clustering algorithm based on unsupervised learning is proposed to objectively classify the fatigue status.Te rhythmic energy and non-linear system features of EEG signals were Extracted through wavelet packet decomposition,and the extracted feature vectors were analyzed and trained using Density Peaks Clustering Algorithm(DPCA)improved by Common-Near-Neighbor(CNN).Then the Bayesian Information Criterion(BIC)was used to determine the clusters.The experimental results show that the accuracy of the improved BDPCA algorithm can reach more than 85%,which can accurately classify different fatigue state levels represented by EEG signals,and achieve the objective definition of brain fatigue state levels.

关 键 词：疲劳等级 脑电信号 小波包分解 密度峰值聚类 贝叶斯准则 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]