基于多视图多特征集成模型的EEG睡眠分期算法

打开文本图片集

关键词：睡眠分期；集成模型；多视图融合网络；堆叠泛化；自适应增强中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2025）12-021-3691-07doi：10.19734/j.issn. 1001-3695.2025.05.0130

EEG sleep staging algorithm based on multi-view and multi-feature integration model

Wang Xucan 1a，2 ， Zhou Qiang 1a，2† ， Li Wan1b.2 （1.a.choolei Technology，Xi’an710021，China；2.Shaanxi Artificial IntellgenceJointLaboratory，Xi'an71021，China）

Abstract：With single-channel electroencephalogram （EEG）signalanalysis replacing multi-channel EEGas the main sleep stagingmode，thespatialinformationinthoriginaltime-space-frequencythre-dimensionalinformationislostndmostsleep stagingmethodslackfrequencyanalysiscapabilities，resultingintheunderutilizationof theEEGfrequencydomain.Inaddition，the classimbalance among sleepstages has jointlyledtothediffculty inimproving the acuracyof sleepstaging methods.Aimingat theaboveshortcomings，this paper proposedasleepstagingalgorithmIM-MVFNetbasedonintegrated model （IM）admulti-viewfusionnetwork（MVFNet）.Inorder tofullytapthediffrentdimensional informationofEEGsignals，it respectivelyusedone-dimensionalandtwo-dimensionalnetworkstoextractfeaturesfromtheoriginalEEGsignalsinthetimespace domainandtheshort-time powerspectrummapsofEEGinthetime-frequencydomain.Then，itusedthestackinggeneralization（SG）method tocombinethetwoheterogeneous networks tofusetheextracteddiversefeatures.Finalyitusedthe adaptive bosting（AdaBoost）algorithm to complete thesleep staging.The proposed IM-MVFNet achievesanaccuracyof （204号 89.6% on the Fpz-Cz and Pz-Oz channels of the public dataset Sleep-EDFx，and various indicators are better than other algorithms in recent years.

Key words：sleep staging；integration model；multi view fusion network；stack generalization；adaptive boosting

0 引言

睡眠与人体健康密切相关，充足的睡眠在提高认知能力、调节情绪和增强免疫系统功能等方面起着至关重要的作用[1]。（剩余18607字）