Page - 319 - in Intelligent Environments 2019 - Workshop Proceedings of the 15th International Conference on Intelligent Environments

halationprofile.Whenapatient isusing the inhaler, theadd-ondevicegeneratesasound that is capturedby themicrophoneofa smart-phone.Theacoustic featuresareextracted and thenanalyzedusingsignalprocessingandmachine learningmethods, thusallowing theusers toassess their inhalationandbetter control their condition. In this paper, using theapproachproposed [23],wepresent anautomatedapproach to estimate the inhalation flow rate by analysing the sound generated during the use of a DPI inhaler. The sound signal recorded by the smart-phone microphone to first processedbyabandpassFIRfilter.Thefilteredsignalis thenanalysedtodetect inhalation segments using aBayesian sequential detection algorithm. The energy related to these segments are then computed, and used to estimate lung function parameters, such as the inspiratory flow rate, the inspiratory volume over the first second (FIV1) and the total inspiratory lung volume and others, using linear regression. The proposed system is illustrated in Figure 1. This system can be implemented and integrated in amobile application thatmay help asthma patients to assess their drug administration for better control, and also tomonitor their respiratory function in order to potentially predict the riskofany futureexacerbation.The implementationand the integrationof the systemin amobileapplication isbeyond the scopeof thisworkand isplanned for future studies The remainder of this paper is organized as follows. In section2,we introduceour methodologywhichconsistsof soundacquisition, signalprocessing, segmentationalgo- rithm and inhalation parameters estimation. Section 3 is devoted to estimation results. Finally, section4presentssomeperspectivesfor futureworkanddrawssomeconcluding remarks. Figure1. Acousticmonitoringsystemforasthmapatients 2. Methodology Firstwe startwith data acquisition. Second,wepre-process the acoustic signal through filteringandsegmentation.Theprocessedsignal is thenusedtoextract inhalationmetrics that will be used to estimate the flow rate and other parameters. Finally, we use linear regression toestimate the inhalationparameters fromthecomputedenergy. 2.1. Inhalationsoundsacquisition Theinhalationinducedacousticsignalswererecordedinanoise-freeroomattheUniver- sity ofCopenhagen.Adosage unit sampling apparatus (DUSA)was assembled to per- formtheDPItestasshowninFigure2.Thisunit iscomposedofavacuumpump(HCP5, Z.Jeddi etal. /Estimationof InhalationFlowParameters forAsthmaMonitoring 319