正在加载图片...
第17卷 智能系统学报 ·1060· 法在6个被试上达到了最优。本文特征联合方法 hash[J].Computer engineering science,2021,43(9) 在两个数据集上的均值都达到了最优。 1616-1622 [2] 0.90 毛莺池,唐江红,王静,等.基于Faster R-CNN的多任务 0.85 特征提取方法 YDAWN 增强裂缝图像检测方法円.智能系统学报,2021,16(2) 0.80 切线空间投影 286-293 0.75 直接联合 本文联合 MAO Yingchi,TANG Jianghong,WANG Jing,et al. 盆0.70 0.65 Multi-task enhanced dam crack image detection based on 0.60 Faster R-CNN[J].CAAI transactions on intelligent sys- 0.55 tems,2021,16(2:286-293. 0.50 1 2 3 4567891011均值 [3] 丁斌.基于非线性哈希的图像与视频检索算法研究 被试编号 D].北京:北京邮电大学,2020 图9 PhysioNetRSVP不同被试的平衡准确率 DING Bin.Research on image and video retrieval al- Fig.9 BCA of different subjects under PhysioNetRSVP gorithm based on nonlinear hashing[D].Beijing:Beijing 0.95 University of Posts and Telecommunications,2020. 特征提取方法 0.90 XDAWN [4] WOODMAN G F.A brief introduction to the use of 0.85 一切线空间投影 0.80 event-related potentials in studies of perception and atten- s0.75 tion[J].Attention,perception psychophysics,2010, 立0.70 72(8):2031-3046. 0.65 0.60 [5]LIU Shuang,WANG Wei,SHENG Yue,et al.Improving 0.55 the cross-subject performance of the ERP-based brain- 0.50 2 3 4 567 8910均值 computer interface using rapid serial visual presentation 被试编号 and correlation analysis rank[J.Frontiers in human neur- 图10清华RSVP不同被试的平衡准确率 oscience,2020,14-296. Fig.10 BCA of different subjects under Tsinghua RSVP [6] SOLIS-ESCALANTE T,GENTILETTI GG,YANEZ- SUAREZ O.Single trial P300 detection based on the em- 4结束语 pirical mode decomposition[C]//2006 International Con- 本文提出了一种面向跨被试RSVP的多特征 ference of the IEEE Engineering in Medicine and Bio 低维子空间嵌入的ERP检测方法。采用欧式空 logy Society.New York,IEEE,2006:1157-1160. 间对齐作为迁移方法,平衡准确率作为评价指 [7] KRUSIENSKI D J.SELLERS E W.MCFARLAND D J. 标,留一被试法作为检验方法,分别在PhysioN- et al.Toward enhanced P300 speller performance[J]. etRSVP数据集和清华RSVP数据集下,探索了该 Journal of neuroscience methods,2008,167(1):15-21. [8] RIVET B,SOULOUMIAC A,ATTINA V,et al. 方法在不同脑电长度分段以及不同被试下的表 xDAWN algorithm to enhance evoked potentials:applica- 现,并且与切线空间投影特征提取方法、xDAWN tion to brain-computer interface[J].IEEE transactions on 特征提取方法以及直接特征联合方法进行了对 bio-medical engineering,2009,56(8):2035-2043 比。本文特征联合方法在两个数据集共计14个 [9] RIVET B.CECOTTI H.SOULOUMIAC A.et al.Theor- 长度分段下,有12个长度分段达到最优分类效 etical analysis of xDAWN algorithm:application to an ef- 果。在两个数据集的超过半数被试上达到了最优 ficient sensor selection in a p300 BCI[C]//2011 19th 的分类效果。实验结果表明,本文提出的特征联 European Signal Processing Conference.Barcelona, 合方法能够有效整合来自两个不同空间的特征, EEE,2011:1382-1386. 使得分类结果更具可靠性。 [10]LAWHERN V J,SOLON A J,WAYTOWICH N R,et al. 参考文献: EEGNet:a compact convolutional neural network for EEG-based brain-computer interfaces[J].Journal of neur- [1]周燕,潘丽丽,陈蓉玉,等.结合自适应融合网络与哈希 al engineering,2018,15(5)-056013 的图像检索算法[U.计算机工程与科学,2021,43(9): [11]ZANINI P,CONGEDO M,JUTTEN C,et al.Transfer 1616-1622 learning:a Riemannian geometry framework with applic- ZHOU Yan,PAN Lili,CHEN Rongyu,et al.A novel im- ations to brain-computer interfaces[J].IEEE transactions age retrieval algorithm with adaptive fusion network and on biomedical engineering,2018,65(5):1107-1116.法在 6 个被试上达到了最优。本文特征联合方法 在两个数据集上的均值都达到了最优。 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 BCA 1 2 3 4 5 6 7 8 9 10 11均值 被试编号 特征提取方法 xDAWN 切线空间投影 直接联合 本文联合 图 9 PhysioNetRSVP 不同被试的平衡准确率 Fig. 9 BCA of different subjects under PhysioNetRSVP 0.90 0.95 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 BCA 特征提取方法 xDAWN 切线空间投影 直接联合 本文联合 1 2 3 4 5 6 7 8 9 10 均值 被试编号 图 10 清华 RSVP 不同被试的平衡准确率 Fig. 10 BCA of different subjects under Tsinghua RSVP 4 结束语 本文提出了一种面向跨被试 RSVP 的多特征 低维子空间嵌入的 ERP 检测方法。采用欧式空 间对齐作为迁移方法,平衡准确率作为评价指 标,留一被试法作为检验方法,分别在 PhysioN￾etRSVP 数据集和清华 RSVP 数据集下,探索了该 方法在不同脑电长度分段以及不同被试下的表 现,并且与切线空间投影特征提取方法、xDAWN 特征提取方法以及直接特征联合方法进行了对 比。本文特征联合方法在两个数据集共计 14 个 长度分段下,有 12 个长度分段达到最优分类效 果。在两个数据集的超过半数被试上达到了最优 的分类效果。实验结果表明,本文提出的特征联 合方法能够有效整合来自两个不同空间的特征, 使得分类结果更具可靠性。 参考文献: 周燕, 潘丽丽, 陈蓉玉, 等. 结合自适应融合网络与哈希 的图像检索算法 [J]. 计算机工程与科学, 2021, 43(9): 1616–1622. ZHOU Yan, PAN Lili, CHEN Rongyu, et al. A novel im￾age retrieval algorithm with adaptive fusion network and [1] hash[J]. Computer engineering & science, 2021, 43(9): 1616–1622. 毛莺池, 唐江红, 王静, 等. 基于 Faster R-CNN 的多任务 增强裂缝图像检测方法 [J]. 智能系统学报, 2021, 16(2): 286–293. MAO Yingchi, TANG Jianghong, WANG Jing, et al. Multi-task enhanced dam crack image detection based on Faster R-CNN[J]. CAAI transactions on intelligent sys￾tems, 2021, 16(2): 286–293. [2] 丁斌. 基于非线性哈希的图像与视频检索算法研究 [D]. 北京: 北京邮电大学, 2020. DING Bin. Research on image and video retrieval al￾gorithm based on nonlinear hashing[D]. Beijing: Beijing University of Posts and Telecommunications, 2020. [3] WOODMAN G F. A brief introduction to the use of event-related potentials in studies of perception and atten￾tion[J]. Attention, perception & psychophysics, 2010, 72(8): 2031–3046. [4] LIU Shuang, WANG Wei, SHENG Yue, et al. Improving the cross-subject performance of the ERP-based brain￾computer interface using rapid serial visual presentation and correlation analysis rank[J]. Frontiers in human neur￾oscience, 2020, 14–296. [5] SOLIS-ESCALANTE T, GENTILETTI G G, YANEZ￾SUAREZ O. Single trial P300 detection based on the em￾pirical mode decomposition[C]//2006 International Con￾ference of the IEEE Engineering in Medicine and Bio￾logy Society. New York, IEEE, 2006: 1157−1160. [6] KRUSIENSKI D J, SELLERS E W, MCFARLAND D J, et al. Toward enhanced P300 speller performance[J]. Journal of neuroscience methods, 2008, 167(1): 15–21. [7] RIVET B, SOULOUMIAC A, ATTINA V, et al. xDAWN algorithm to enhance evoked potentials: applica￾tion to brain-computer interface[J]. IEEE transactions on bio-medical engineering, 2009, 56(8): 2035–2043. [8] RIVET B, CECOTTI H, SOULOUMIAC A, et al. Theor￾etical analysis of xDAWN algorithm: application to an ef￾ficient sensor selection in a p300 BCI[C]//2011 19th European Signal Processing Conference. Barcelona, IEEE, 2011: 1382−1386. [9] LAWHERN V J, SOLON A J, WAYTOWICH N R, et al. EEGNet: a compact convolutional neural network for EEG-based brain-computer interfaces[J]. Journal of neur￾al engineering, 2018, 15(5)–056013. [10] ZANINI P, CONGEDO M, JUTTEN C, et al. Transfer learning: a Riemannian geometry framework with applic￾ations to brain–computer interfaces[J]. IEEE transactions on biomedical engineering, 2018, 65(5): 1107–1116. [11] 第 17 卷 智 能 系 统 学 报 ·1060·
<<向上翻页向下翻页>>
©2008-现在 cucdc.com 高等教育资讯网 版权所有