Wireless Brain-to-Computer Connection Synthesizes Speech

《Wireless Brain-to-Computer Connection Synthesizes Speech》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、AWirelessBrain-MachineInterfaceforReal-TimeSpeechSynthesis1,21,334FrankH.Guenther*,JonathanS.Brumberg,E.JosephWright,AlfonsoNieto-Castanon,JasonA.111333,5Tourville,MikhailPanko,RobertLaw,StevenA.Siebert,JessL.Bartels,DinalS.Andreasen,673PrincewillEhirim,HuiMao,PhilipR.Kennedy1DepartmentofCognitive

2、andNeuralSystemsandSargentCollegeofHealthandRehabilitationSciences,BostonUniversity,Boston,Massachusetts,UnitedStatesofAmerica,2DivisionofHealthSciencesandTechnology,HarvardUniversity-MassachusettsInstituteofTechnology,Cambridge,Massachusetts,UnitedStatesofAmerica,3NeuralSignalsInc.,Duluth,Georgia

3、,UnitedStatesofAmerica,4StatsANCLLC,BuenosAires,Argentina,5GeorgiaTechResearchInstitute,Marietta,Georgia,UnitedStatesofAmerica,6GwinnettMedicalCenter,Lawrenceville,Georgia,UnitedStatesofAmerica,7EmoryCenterforSystemsImaging,EmoryUniversityHospital,Atlanta,Georgia,UnitedStatesofAmericaAbstractBackg

4、round:Brain-machineinterfaces(BMIs)involvingelectrodesimplantedintothehumancerebralcortexhaverecentlybeendevelopedinanattempttorestorefunctiontoprofoundlyparalyzedindividuals.CurrentBMIsforrestoringcommunicationcanprovideimportantcapabilitiesviaatypingprocess,butunfortunatelytheyareonlycapableofsl

5、owcommunicationrates.Inthecurrentstudyweuseanovelapproachtospeechrestorationinwhichwedecodecontinuousauditoryparametersforareal-timespeechsynthesizerfromneuronalactivityinmotorcortexduringattemptedspeech.Methodology/PrincipalFindings:NeuralsignalsrecordedbyaNeurotrophicElectrodeimplantedinaspeech-

6、relatedregionoftheleftprecentralgyrusofahumanvolunteersufferingfromlocked-insyndrome,characterizedbynear-totalparalysiswithsparedcognition,weretransmittedwirelesslyacrossthescalpandusedtodriveaspeechsynthesizer.AKalmanfilter-baseddecodertranslatedtheneuralsignalsgeneratedduringattemptedspeechintoc

7、ontinuousparametersforcontrollingasynthesizerthatprovidedimmediate(within50ms)auditoryfeedbackofthedecodedsound.Accuracyofthevolunteer’svowelproductionswiththesynthesizerimprovedquicklywithpractice,witha25%improv