• DocumentCode
    1824724
  • Title

    iKnow Where You Are

  • Author

    Subbu, Kalyan ; Xu, Ning ; Dantu, Ram

  • Author_Institution
    Dept. of Comput. Sci. & Eng., Univ. of North Texa, Denton, TX, USA
  • Volume
    4
  • fYear
    2009
  • fDate
    29-31 Aug. 2009
  • Firstpage
    469
  • Lastpage
    474
  • Abstract
    ldquoSmartrdquo phones, such as G1 and iPhone, have made their way into people´s lives with more intelligence owing to the continually decreasing cost and increasing access to memory capability, processing power and network bandwidth. This work attempts to detect presence in different environments like office, conference, meeting and traveling outdoors, using audio sensors on the G1 phone. Specifically, a two step process involving classifying the background first and then detecting the number of people in an environment is followed. Using the Vector Quantization method, audio is classified into five predefined classes. A recognition rate ranging from 86% to 100%for individual classes and 91%overall recognition accuracy was obtained. For speaker change detection via Bayesian Information Criterion, about 79.4% of all 350 test audio clips was correctly categorized.
  • Keywords
    Bayes methods; audio signal processing; mobile computing; signal classification; speaker recognition; vector quantisation; Bayesian information criterion; G1 phone; audio classification; audio sensors; background classification; iPhone; memory capability; network bandwidth; presence detection; processing power; smart phones; speaker change detection; speaker recognition; vector quantization; Acoustic signal detection; Bandwidth; Costs; Intelligent networks; Intelligent sensors; Layout; Loudspeakers; Smart phones; Speech; Working environment noise;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Computational Science and Engineering, 2009. CSE '09. International Conference on
  • Conference_Location
    Vancouver, BC
  • Print_ISBN
    978-1-4244-5334-4
  • Electronic_ISBN
    978-0-7695-3823-5
  • Type

    conf

  • DOI
    10.1109/CSE.2009.55
  • Filename
    5284201