| 周飞扬*,柳政卿*,王秋成*,杨 忠**.基于拓扑数据分析的驾驶疲劳EEG 数据处理与优化分析研究[J].高技术通讯(中文),2023,33(3):322~331 |
| 基于拓扑数据分析的驾驶疲劳EEG 数据处理与优化分析研究 |
| Analysis of driving fatigue EEG signals based on topological data analysis |
| |
| DOI：10. 3772/ j. issn. 1002-0470. 2023. 03. 011 |
| 中文关键词: 疲劳驾驶;脑电信号(EEG);拓扑数据分析(TDA);持久同源(PH);支持向量机(SVM) |
| 英文关键词: fatigue driving, electroencephalogram(EEG)signal, topological data analysis(TDA), persistenthomology(PH), support vector machine(SVM) |
| 基金项目: |
| 作者 | 单位 | | 周飞扬* | (*浙江工业大学机械工程学院 杭州310023)(**深蓝汽车科技有限公司 重庆400020) | | 柳政卿* | (*浙江工业大学机械工程学院 杭州310023)(**深蓝汽车科技有限公司 重庆400020) | | 王秋成* | (*浙江工业大学机械工程学院 杭州310023)(**深蓝汽车科技有限公司 重庆400020) | | 杨 忠** | (*浙江工业大学机械工程学院 杭州310023)(**深蓝汽车科技有限公司 重庆400020) |
|
| 摘要点击次数: 924 |
| 全文下载次数: 750 |
| 中文摘要: |
| 为提高驾驶疲劳脑电(EEG)数据处理与分析的准确性和鲁棒性,提出一种基于拓扑数据分析(TDA)的驾驶人疲劳脑电分析方法。首先利用汽车性能虚拟仿真平台开展驾驶实验,通过驾驶人状态反馈和面部特征视频,标记脑电数据,形成清醒和疲劳二分数据集。之后利用EEGLAB 预处理数据,剔除噪声并保留0.3 ~30 Hz 频带,直接从时域EEG 数据中提取拓扑特征。此外还提取了经典频域特征α 波能量和α/ β 用于对比分析。最后使用支持向量机进行分类。结果表明,基于持久同源(PH)的拓扑特征取得了高达88.7%的准确率和91.4%的召回率,与经典频域特征性能相当,且对脑电伪影的鲁棒性明显更好,在未剔除EEG 伪影的情况下仍取得了87.4%的准确率和89.7%的召回率。综上所述,本文提出的用于驾驶疲劳脑电信号处理与分析的TDA 方法抗干扰特性好、处理成本低、经济性高,有助于稳定、高效地处理驾驶人脑电数据并检测驾驶疲劳状态,具有较大的科学实际应用价值。 |
| 英文摘要: |
| In this paper, topological data analysis(TDA)is used to enhance the accuracy and robustness of driving fatigue Electroencephalogram(EEG) signals processing and data analysis. The experiment of occupant driving fatigue is conducted by using a driving simulator. The experimental obtained EEG data are labeled by combining the fatigue self-evaluation results and recorded driver??s facial video, to form a binary data set of wake-up and fatigue. The EEG data are preprocessed by using EEGLAB, and the noises are carefully eliminated and a 0. 3 ~30 Hz frequency band is retained. The classical frequency domain characteristics α wave energy and α/ β are extracted for comparative analysis and research. Finally, the extracted features are classified by support vector machine. The results show that the accuracy of 88. 7% and the recall rate of 91. 4% are obtained based on the topological characteristics of persistent homology(PH). The performance agrees well with the classical frequency domain features. The topological features are significantly more robust to EEG artifacts, and the accuracy of 87. 4% and the recall rate of 89. 7% are achieved in the EEG data which has not eliminated artifacts. The proposed TDA method for driving fatigue EEG signals process and analysis has good anti-interference characteristics, low data processing cost, and high economy.It is helpful for stable and efficient detection of driver fatigue state and has a great scientific-practical application value. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
