运动想象脑电信号的识别与分类问题一直是脑机领域研究的热点问题。针对此问题,使用区别传统线性降维方法的流形学习方法,将共空间模式算法与均匀流形投影算法相结合,充分利用了脑电信号中的非线性特征,对运动想象脑电信号进行了特征提取和数据降维,并使用KNN分类器进行了分类,对分类效果做出了评价;将降维前后的数据分类结果进行对比,说明了数据降维的优点和必要性;进一步讨论了降维结果在数据可视化方面的表现。发现经过数据降维的特征数据的可视化效果明显优于未经过降维的数据,进一步提出了一种基于共空间模式和均匀流形投影的新型脑电信号识别方法,对进行脑电信号深度剖析。挖掘脑电信号非线性特征提供了参考价值,同时也在数据流形分布以及数据可视化的角度为运动想象脑电信号识别提供了新思路。
Abstract
The recognition and classification of motor imaging EEG signals has always been a hot issue in the field of brain-computer research. In response to this problem, uses a manifold learning method that is different from the traditional linear dimensionality reduction method, combine the CSP and UMAP, make full use of the non-linear features in the EEG signal, perform feature extraction and data dimensionality reduction on the motor imagination EEG signal, and use the KNN classifier to classify, and the classification effect is evaluated; The comparison of the data classification results before and after dimensionality reduction illustrates the advantages and necessity of data dimensionality reduction; The performance of dimensionality reduction results in data visualization is further discussed, and it is found that the visualization effect of feature data after data dimensionality reduction is significantly better than that of data without dimensionality reduction. A new EEG signal recognition method based on CSP and UMAP is proposed, which provides reference value for deep analysis of EEG signals and mining of EEG nonlinear characteristics. The angle of data visualization provides new ideas for the recognition of EEG signals in motor imagination.
关键词
计量学 /
脑电信号 /
运动想象 /
流形学习 /
均匀流行投影 /
共空间模式 /
数据降维
Key words
metrology /
EEG signal /
motor imagination /
manifold learning /
UMAP /
CSP /
data dimensionality reduction
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
［1］何群, 杜硕, 王煜文, 等. 基于变分模态分解与深度信念网络的运动想象分类识别研究 ［J］. 计量学报, 2020, 41 (1): 90-99.
He Q, Du S, Wang Y W, et al. The Classification of EEG Induced by Motor Imagery Based on Variational Mo-de Decomposition and Deep Belief Network ［J］. Acta Metrologica Sinica, 2020, 41 (1): 90-99.
［2］樊凤杰, 白洋, 纪会芳. 基于EEMD-ICA的脑电去噪算法研究 ［J］. 计量学报, 2021, 42 (3): 395-400.
Fan F J, Bai Y, Ji H F. Denoising Method of EEG Signal Based on EEMD-ICA ［J］. Acta Metrologica Sinica, 2021, 42 (3): 395-400.
［3］朱建新, 吕宝林, 乔松, 等. 基于主成分分析及多维高斯贝叶斯的超声流量计故障智能诊断方法 ［J］. 计量学报, 2020, 41 (12): 1494-1499.
Zhu J X, Lv B L, Qiao S, et al. Application of Primary Component Analysis and Multivariate Gaussian Bayesian Method on Intelligent Failure Diagnosis of Ultrasonic Flo-wmeter, ［J］. Acta Metrologica Sinica, 2020, 41 (12): 1494-1499.
［4］Seung H S, Lee D D. The manifold ways of perception ［J］. Science, 2000, 290 (5500): 2268-2269.
［5］汤宝平, 马婧华. 多准则融合敏感特征选择和自适应邻域的流形学习故障诊断 ［J］. 仪器仪表学报, 2014, 35 (11): 2415-2422.
Tang B P, Ma J H. Manifold learning method for fault diagnosis based on sensitive feature selection with multi-criteria evaluation sequences and adaptive neighborhood ［J］. Chinese Journal of Scientific Instrument, 2014, 35 (11): 2415-2422.
［6］Vaughan T M. Editorial brain-computer interface techno-logy: A review of the second international meeting ［J］. IEEE Transaction on Neural Systems and Rehabilitation Engineering, 2015, 11 (2): 94-109.
［7］Hong S, Zhang J, Li T, et al. A manifold learning approach to urban land cover classification with optical and radar data ［J］. Landscape and Urban Planning, 2018, 172: 11-24.
［8］黄伟, 刘战民, 薛丹, 等. 基于流形学习的聚类方法在基因芯片表达谱分析中的应用 ［J］. 中国生物医学工程学报, 2010, 29 (1): 77-85.
Huang W, Liu Z M, Xue D, et al. The Application of Cluster Methods Based on Manifold Leaning in Analysis of Gene Expression Profile ［J］. Chinese Journal of Biom-edical Engineering, 2010, 29 (1): 77-85.
［9］Na Q, Yong K, Sun P J, et al. Bogie Fault Identification Based on EEMD Information Entropy and Manifold Lear-ning ［J］. IFAC PapersOnLine, 2017, 50 (1): 315-318.
［10］Barsotti M, Leonardis D, Loconsole C, et al. A full upper limb robotic exoskeleton for reaching and grasping rehabilitation triggered by MI-BCI［C］// IEEE Intern-ational Conference on Rehabilitation Robotics. IEEE, 2015: 49-54.
［11］Rutkowski T M, Mandic D P, Cichocki A, et al. EMD Approach to Multichannel EEG Data-The Amplitude and Phase Synchrony Analysis Technique ［J］. Journal of Circuits, Systems, and Computers, 2010, 19 (1): 215-229.
［12］Arvaneh M, Guan C, Ang K A, et al. Spatially sparsed Common Spatial Pattern to improve BCI performance［C］//IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2011: 2412-2424.
［13］Shi L C, Li Y, Sun R H, et al. A sparse common spatial pattern algorithm for brain-computer interface ［J］. Neural Inf Process, 2011, 7062: 725-733.
［14］Blankertz B, Muller K R, Krusienski D J, et al. The BCI competition III: Validating alternative approachs to actual BCI problems ［J］. IEEE Trans Neural Sys Re-hab Eng, 2006, 14 (2): 153-159.
［15］Li D L, Xu J C, Wang J H, et al. A Multi-Scale Fusion Convolutional Neural Network Based on Attention Mech-anism for the Visualization Analysis of EEG Signals De-coding ［J］. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2020, 28(12): 2615-2626.
［16］Bosch-Bayard J, Aubert-Vazquez E, Brown S T, et al. A Quantitative EEG Toolbox for the MNI Neuro infor-matics Ecosystem: Normative SPM of EEG Source Spectra ［J］. Frontiers in Neuro informatics, 2020, 14: 33.
［17］樊凤杰,白洋,纪会芳. 基于EEMD-ICA的脑电去噪算法研究［J］. 计量学报, 2021, 42(3): 395-400.
Fan F J, Bai Y, Ji H F. Denoising Method of EEG Signal Based on EEMD-ICA［J］. Acta Metrologica Sinica, 2021, 42(3): 395-400.
［18］付荣荣,米瑞甫,王涵, 等. 基于脑功能网络的脑疲劳状态检测研究［J］. 计量学报, 2021, 42(11): 1528-1533.
Fu R R, Mi R F, Wang H, et al. Research on Fatigue Driving Recognition Based on Brain Function Network［J］. Acta Metrologica Sinica, 2021, 42(11): 1528-1533.
基金
国家自然科学基金(62073282);河北省中央引导地方科技发展资金(206Z0301G);河北省自然科学基金(E2018203433);河北省教育厅在读研究生创新能力培养资助项目(CXZZSS2021060)
PDF(2381 KB)
