?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=DeepEar%3A+Robust+Smartphone+Audio+Sensing+in+Unconstrained+Acoustic+Environments+using+Deep+Learning&rft.creator=Lane%2C+ND&rft.creator=Georgiev%2C+P&rft.creator=Qendro%2C+L&rft.description=Microphones+are+remarkably+powerful+sensors+of+human+behavior+and+context.+However%2C+audio+sensing+is+highly+susceptible+to+wild+fluctuations+in+accuracy+when+used+in+diverse+acoustic+environments+(such+as%2C+bedrooms%2C+vehicles%2C+or+cafes)%2C+that+users+encounter+on+a+daily+basis.+Towards+addressing+this+challenge%2C+we+turn+to+the+field+of+deep+learning%3B+an+area+of+machine+learning+that+has+radically+changed+related+audio+modeling+domains+like+speech+recognition.+In+this+paper%2C+we+present+DeepEar+-+the+first+mobile+audio+sensing+framework+built+from+coupled+Deep+Neural+Networks+(DNNs)+that+simultaneously+perform+common+audio+sensing+tasks.+We+train+DeepEar+with+a+large-scale+dataset+including+unlabeled+data+from+168+place+visits.+The+resulting+learned+model%2C+involving+2.3M+parameters%2C+enables+DeepEar+to+significantly+increase+inference+robustness+to+background+noise+beyond+conventional+approaches+present+in+mobile+devices.+Finally%2C+we+show+DeepEar+is+feasible+for+smartphones+by+building+a+cloud-free+DSP-based+prototype+that+runs+continuously%2C+using+only+6%25+of+the+smartphone's+battery+daily.&rft.subject=Mobile+sensing%2C+deep+learning%2C+audio+sensing.&rft.publisher=Association+for+Computing+Machinery&rft.contributor=Miyahara%2C+H&rft.contributor=Tokuda%2C+H&rft.contributor=Mase%2C+K&rft.contributor=Langheinrich%2C+M&rft.date=2015-09-11&rft.type=Proceedings+paper&rft.language=eng&rft.source=+++++In%3A+Miyahara%2C+H+and+Tokuda%2C+H+and+Mase%2C+K+and+Langheinrich%2C+M%2C+(eds.)+Proceedings+of+the+2015+ACM+International+Joint+Conference+on+Pervasive+and+Ubiquitous+Computing.++(pp.+pp.+283-294).++Association+for+Computing+Machinery%3A+Osaka%2C+Japan.+(2015)+++++&rft.format=text&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F1503675%2F1%2FLane%252C%2520Georgiev%252C%2520%2526%2520Qendro%25202015%2520UbiComp.pdf&rft.identifier=https%3A%2F%2Fdiscovery.ucl.ac.uk%2Fid%2Feprint%2F1503675%2F&rft.rights=open