Page - 375 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

The lackof largedatasetspenalizes thepossibilityof exploitingdeep learning tech- niques that requirea lotofdatabutprovidingaverygoodaccuracy. The goal of this work is to propose a platformwhich firstly integrates data from heterogenous sourcesandsecondlyprovides several typesofaccess to thedata. The framework has been partially implemented.We have prioritized the develop- ment of themost challengingmodules:DataCollection andDataManagement compo- nents.Till now,fivedatasetshavebeen integrated. Futuredirectionsincludethedevelopmentoftheremainingmodulesandanintensive testof theoverallplatform.Oncethe testsarecompleted,wewillmakeavailable the tool and the specifications for thedesignof themodule for importingnewdatasets. References [1] Sensor data logger. https://play.google.com/store/apps/details?id=net.steppschuh. sensordatalogger, accessedon22March2019. [2] DavideAnguita, AlessandroGhio, LucaOneto, Xavier Parra, and Jorge Luis Reyes-Ortiz. A public domaindataset forhumanactivity recognitionusingsmartphones. InProceedingsof theEuropeanSym- posiumonArtificialNeuralNetworks,Computational IntelligenceandMachineLearning (ESANN13), 2013. [3] FabioBagala, ClemensBecker, AngeloCappello, LorenzoChiari, KamiarAminian, JeffreyMHaus- dorff,WiebrenZijlstra, and JochenKlenk. Evaluationof accelerometer-based fall detectionalgorithms on real-world falls.PloSone, 7(5):e37062,2012. [4] Tadas Baltrusˇaitis, Chaitanya Ahuja, and Louis-PhilippeMorency. Multimodal machine learning: A surveyandtaxonomy. IEEETransactionsonPatternAnalysisandMachineIntelligence,41(2):423–443, 2019. [5] JamesBartlett, Vinay Prabhu, and JohnWhaley. Acctionnet:A dataset of human activity recognition usingon-phonemotion sensors. InProceedings of the InternationalConferenceonMachineLearning (ICML17), 2017. [6] YoshuaBengio. Deep learning of representations: Looking forward. In International Conference on StatisticalLanguageandSpeechProcessing (SLSP13), 2013. [7] SimoneBianco,RemiCadene,LuigiCelona, andPaoloNapoletano. Benchmark analysis of represen- tativedeepneuralnetworkarchitectures. IEEEAccess, 6:64270–64277,2018. [8] AndreasBulling,UlfBlanke, andBernt Schiele. A tutorial onhumanactivity recognitionusingbody- worn inertial sensors.ACMComputingSurveys (CSUR), 46(3):33,2014. [9] EduardoCasilari, JoseASantoyo-Ramo´n,andJoseMCano-Garcı´a. Umafall:Amultisensordataset for the researchonautomatic fall detection.ProcediaComputerScience, 110:32–39,2017. [10] M.Cornacchia,K.Ozcan,Y.Zheng,andS.Velipasalar.Asurveyonactivitydetectionandclassification usingwearable sensors. IEEESensorsJournal, 17(2):386–403,2017. [11] AnnaFerrari,DanielaMicucci,MarcoMobilio, andPaoloNapoletano. On thepersonalizationof clas- sificationmodels forhumanactivity recognition. IEEEAccess, submitted. [12] Anna Ferrari, Daniela Micucci, Marco Mobilio, and Paolo Napoletano. On the homogenization of heterogeneous inertial-baseddatabases forhumanactivity recognition. InSubmitted to the IEEEWorld CongressonServices,WorkshoponBigData forPublicHealthPolicyMaking, 2019. [13] DavideGinelli,DanielaMicucci,MarcoMobilio,andPaoloNapoletano.Unimibaal:Anandroidsensor dataacquisitionand labelingsuite.AppliedSciences, 8(8), 2018. [14] DannyHarnik,BennyPinkas,andAlexandraShulman-Peleg. Sidechannels incloudservices:Dedupli- cation incloudstorage. IEEESecurity&Privacy, 8(6):40–47,2010. [15] Andrew P. Hills, Najat Mokhtar, and Nuala M. Byrne. Assessment of physical activity and energy expenditure:Anoverviewofobjectivemeasures.Frontiers inNutrition, 1(5), 2014. [16] ChristianKrupitzer,TimoSztyler, JanickEdinger,MartinBreitbach,HeinerStuckenschmidt,andChris- tianBecker. Hipsdo lie! aposition-awaremobile fall detection system. In IEEEInternationalConfer- enceonPervasiveComputingandCommunications (PerCom18), 2018. [17] NicholasDLane,YeXu,HongLu,ShaohanHu,TanzeemChoudhury,AndrewTCampbell, andFeng Zhao. Enabling large-scale human activity inference on smartphones using community similarity net- A.Ferrari etal. /AFramework forLong-TermDataCollection 375