首页 | 官方网站   微博 | 高级检索  
     

基于脑功能网络深度学习的车内噪声评价模型
引用本文:邹丽媛,王宏,宋桂秋.基于脑功能网络深度学习的车内噪声评价模型[J].机械与电子,2020,38(5):76-80.
作者姓名:邹丽媛  王宏  宋桂秋
作者单位:1.东北大学机械工程与自动化学院,辽宁 沈阳 110819;2.辽宁水利职业学院,辽宁 沈阳 110122
基金项目:国家重点研发计划;辽宁省高等学校创新团队项目
摘    要:研究并构建了一个结合脑电信号处理与深度学习的车内噪声评价模型,该算法通过自我学习实现脑电信号特征提取,使用同步似然方法构建delta、alpha和beta频段的脑功能网络。将3个频带的脑功能网络扁平化处理后作为输入,通过无监督的堆栈自编码器(RSAE)自主提取脑功能网络的特征。通过几个高阶特征训练前后对比,证实了RSAE自主学习到与噪声评价有关的脑神经特征。最终将RSAE与普遍使用的SVM回归模型进行比较,同时将脑功能网络与传统的基于心理声学声音品质的车内噪声评价进行对比。结果表现,所提出的脑功能网络RSAE模型的平均决定系数高达98.69%,明显优于其他方法。

关 键 词:车内噪声评价  堆栈自编码器  脑功能网络分析  同步似然

Research on Vehicle Interior Noise Evaluation Model Based on Stacked Antoencoder and Functional Brain Network
ZOULiyuan,' target="_blank" rel="external">,WANGHong,SONGGuiqiu.Research on Vehicle Interior Noise Evaluation Model Based on Stacked Antoencoder and Functional Brain Network[J].Machinery & Electronics,2020,38(5):76-80.
Authors:ZOULiyuan  " target="_blank">' target="_blank" rel="external">  WANGHong  SONGGuiqiu
Affiliation:1.SchoolofMechanicalEngineeringandAutomation,NortheasternUniversity,Shenyang110819,China; 2.LiaoningWaterConservancyVocationalCollege,Shenyang110122,China
Abstract:Inthispaper,anin vehiclenoiseevaluation modelcombiningEEGsignalprocessinganddeep learningwasstudiedandconstructed.ThisalgorithmcanrealizeEEGsignalfeatureextractionthroughself learning.Thesynchronouslikelihoodmethodwasusedtoconstructthebrainfunctionnetworkofdelta,alphaandbeta bands.FortheunsteadycharacteristicsofEEGsignals,thesynchronouslikelihoodmethoddoesbetterinfindingthe linearandnonlinearcouplingrelationshipbetweendifferentchannels.Thethreefrequencybandsoffunctionalbrain networkswereflattenedasaninput.Firstly,thefeaturesofthefunctionalbrainnetworkwereextractedthrough unsupervisedpre training,andthentheinitializationweightswereusedtotrainthefour layernetworkusingthe in vehiclenoisescoresevaluatedbytheexpertgroupastheannotationinformation.Throughthecomparisonof severalhigh orderfeaturesbeforeandaftertraining,thevalidityofRSAE’sself learnedfeatureswereconfirmed tobeuseful.Intheend,thispapercomparedtheRSAEwithSVMregressionmodel,andcomparedthefunctional brainnetworkwiththetraditionalpsychoacousticsoundquality basedinteriornoiseevaluation.Theresultsshow thattheproposedvehicleinteriornoisebasedonRSAEandfunctionalbrainnetworkwasbetter.TheaveragedecisioncoefficientofthefunctionalbrainnetworkRSAEmodelproposedisashighas98.69%,whichisobviouslysuperiortoothermethods.Theresultsalsoconfirmedthefeasibilityandeffectivenessoftheproposedmodel.
Keywords:vehicle interior noise evaluation  stack AutoEncoder  brain network analysis  synchronouslikelihood
本文献已被 CNKI 维普 万方数据 等数据库收录!
点击此处可从《机械与电子》浏览原始摘要信息
点击此处可从《机械与电子》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号