Upgrade to Pro — share decks privately, control downloads, hide ads and more …

情報通信技術を用いたセンシング / Real world sensing with ICT

情報通信技術を用いたセンシング / Real world sensing with ICT

九州大学高等研究院 / 九州先端科学技術研究所 研究交流会 講演資料
Material presented at Kyushu Univ - ISIT Joint Workshop

Shigemi ISHIDA

February 07, 2019
Tweet

More Decks by Shigemi ISHIDA

Other Decks in Technology

Transcript

  1. ৘ใ௨৴ٕज़Λ༻͍ͨ

    ηϯγϯά
    ੴా ൟາ
    ۝भେֶ γεςϜ৘ใՊֶݚڀӃ ॿڭ
    Feb 7, 2019 @۝भେֶ–ISITݚڀަྲྀձ

    View full-size slide

  2. ࢿྉ͸ҎԼͰऔಘͰ͖·͢
    https://speakerdeck.com/
    pman0214/real-world-
    sensing-with-ict

    View full-size slide

  3. ੴా ൟາʢ͍ͩ͠ ͛͠Έʣ
    ■ ུྺ
    ● 2006೥3݄ ࣳӜ޻ۀେֶ޻ֶ෦ిࢠ޻ֶՊ
    ● 2008೥3݄ ౦ژେֶେֶӃ৽ྖҬ૑੒ՊֶݚڀՊम࢜ྃ
    ● 2008೥4݄ʙ2009೥9݄ (ג) ΞΫςΟε։ൃ৬
    ● 2012೥4݄ʙ2013೥9݄ ֶৼ ಛผݚڀһDC2→PD
    ● 2012೥9݄ ౦ژେֶେֶӃ޻ֶܥݚڀՊത࢜ྃ (޻ത)
    ● 2013೥5݄ʙ9݄ ถϛωιλେ Visi;ng Scholar
    ● 2013೥10݄ʙ ݱ৬
    ■ ઐ໳
    ● ແઢηϯαωοτϫʔΫɼ࣮ۭؒηϯγϯάɼ԰಺ଌҐ
    3

    View full-size slide

  4. IoT: Internet of Things
    ■ ਎ͷճΓͷʮϞϊʯ͕Πϯλʔωοτʹͭͳ͕Δ
    ੈք
    ● Ϟϊ͕ηϯα΍ΞΫνϡΤʔλʹͳΓɼଞͷϞϊ
    ͱ࿈ܞͯ͠ศརͳαʔϏεΛ࣮ݱ
    ● ηϯγϯάσʔλΛղੳͯ͠ΞΫνϡΤʔγϣϯ
    4

    View full-size slide

  5. IoT͸σʔλղੳʁ
    5

    View full-size slide

  6. IoT͸σʔλղੳʁ
    5
    Ϗοάσʔλ
    ਓ޻஌ೳ
    ػցֶश
    ਂ૚ֶश

    View full-size slide

  7. IoT͸σʔλղੳʁ
    5
    σʔλͷऔಘͩͬͯେࣄ
    Ϗοάσʔλ
    ਓ޻஌ೳ
    ػցֶश
    ਂ૚ֶश

    View full-size slide

  8. Ξ΢τϥΠϯ
    ■ ແઢηϯαωοτϫʔΫ
    ■ ԰಺ଌҐٕज़
    ● ҟछແઢΛ࢖ͬͨηϯαଌҐ
    ● νϟωϧ෼཭BLEଌҐ
    ■ ৽͍͠ηϯγϯάٕज़
    ● ं྆ηϯγϯά
    • ԻڹΞϓϩʔν
    • ແઢΞϓϩʔν
    ● ԰֎ਓମηϯγϯά
    6
    ←ίί͕ࠓ೔ͷϝΠϯ

    View full-size slide

  9. Ξ΢τϥΠϯ
    ■ ແઢηϯαωοτϫʔΫ
    ■ ԰಺ଌҐٕज़
    ● ҟछແઢΛ࢖ͬͨηϯαଌҐ
    ● νϟωϧ෼཭BLEଌҐ
    ■ ৽͍͠ηϯγϯάٕज़
    ● ं྆ηϯγϯά
    • ԻڹΞϓϩʔν
    • ແઢΞϓϩʔν
    ● ԰֎ਓମηϯγϯά
    6
    ଌҐٕज़ɼηϯγϯάٕज़Λ
    ۦ͚଍Ͱ঺հ
    ←ίί͕ࠓ೔ͷϝΠϯ

    View full-size slide

  10. ແઢηϯαωοτϫʔΫ
    Wireless Sensor Network (WSN)
    7

    View full-size slide

  11. ແઢηϯαωοτϫʔΫ (WSN)
    ■ ηϯαʴແઢ௨৴
    ● औಘͨ͠ηϯασʔλΛແઢ௨৴Ͱऩू
    ■ ྫʣεϚʔτ೶ۀɼεϚʔτϋ΢ε
    ● WSNʹΑΔ؀ڥ৘ใͷऔಘ͕ඞਢ
    8

    View full-size slide

  12. ηϯαϊʔυ
    ■ ௒খܕίϯϐϡʔλʴ௨৴
    ● ۃΊͯඇྗɼػೳ΋গͳ͍
    9
    εϚʔτϑΥϯ
    Moto Z Play
    ηϯαϊʔυ
    MICAz
    CPU 8-core Cortex-A53, 64bit ATmega128L, 8bit
    ΫϩοΫप೾਺ 2.0 GHz 8 MHz
    ROM༰ྔ 32 GB 128 kB
    ϝϞϦ༰ྔ 3 GB 4 kB
    ແઢηϯαϊʔυMICAz

    View full-size slide

  13. ηϯαϊʔυ
    ■ ௒খܕίϯϐϡʔλʴ௨৴
    ● ۃΊͯඇྗɼػೳ΋গͳ͍
    9
    εϚʔτϑΥϯ
    Moto Z Play
    ηϯαϊʔυ
    MICAz
    CPU 8-core Cortex-A53, 64bit ATmega128L, 8bit
    ΫϩοΫप೾਺ 2.0 GHz 8 MHz
    ROM༰ྔ 32 GB 128 kB
    ϝϞϦ༰ྔ 3 GB 4 kB
    ແઢηϯαϊʔυMICAz

    View full-size slide

  14. ηϯαϊʔυ
    ■ ௒খܕίϯϐϡʔλʴ௨৴
    ● ۃΊͯඇྗɼػೳ΋গͳ͍
    9
    εϚʔτϑΥϯ
    Moto Z Play
    ηϯαϊʔυ
    MICAz
    CPU 8-core Cortex-A53, 64bit ATmega128L, 8bit
    ΫϩοΫप೾਺ 2.0 GHz 8 MHz
    ROM༰ྔ 32 GB 128 kB
    ϝϞϦ༰ྔ 3 GB 4 kB
    ͜Εʢ۝େϩΰʣ = 5,619B
    ແઢηϯαϊʔυMICAz

    View full-size slide

  15. ηϯαϊʔυ
    ■ ௒খܕίϯϐϡʔλʴ௨৴
    ● ۃΊͯඇྗɼػೳ΋গͳ͍
    㱺 ͍͔ʹޮ཰ྑ͘࢖͏͔
    9
    εϚʔτϑΥϯ
    Moto Z Play
    ηϯαϊʔυ
    MICAz
    CPU 8-core Cortex-A53, 64bit ATmega128L, 8bit
    ΫϩοΫप೾਺ 2.0 GHz 8 MHz
    ROM༰ྔ 32 GB 128 kB
    ϝϞϦ༰ྔ 3 GB 4 kB
    ͜Εʢ۝େϩΰʣ = 5,619B
    ແઢηϯαϊʔυMICAz

    View full-size slide

  16. WSNʹ͓͚Δ՝୊ͷྫ
    ■ ηϯαϊʔυ͸Ͳ͜ʹઃஔ͞Ε͍ͯΔʁ
    ● શηϯαϊʔυͷҐஔΛखಈͰ؅ཧ͢Δͷ͸΄΅ෆ
    Մೳ
    㱺 ଌҐٕज़ (ಛʹ԰಺͕໰୊)
    ■ ηϯγϯά͕೉͍͠৘ใΛͲ͏΍ͬͯऔಘ͢Δʁ
    ● ैདྷͷηϯαͰ೉͔ͬͨ͜͠ͱΛ࣮ݱ͢Δ৽͍͠η
    ϯγϯάٕज़
    ● লిྗͳηϯγϯάٕज़
    ※ ηϯαϊʔυͰಈ͔͘Ͳ͏͔͸࡞͔ͬͯΒߟ͑Δ
    㱺 ৽͍͠ηϯγϯάٕज़
    10

    View full-size slide

  17. ԰಺ଌҐٕज़
    Indoor LocalizaHon
    11

    View full-size slide

  18. AP1 = –48dBm
    AP2 = –70dBm
    : :
    QUERY
    Estimation
    Phase
    - ଌҐྖҬ֤ॴͰ৴߸ڧ౓Λଌఆ
    - ଌҐ࣌ʹ͸ଌҐ஍఺Ͱଌఆͨ͠
    ৴߸ڧ౓ͱ΋ͬͱ΋ࣅ͍ͯΔ஍
    ఺Λ୳ࡧͯ͠Ґஔਪఆ
    ԰಺ଌҐͷجຊ
    ■ 2ͭͷํࣜ
    12
    DB
    AP1 = –52dBm
    AP2 = –62dBm
    : :
    INSERT
    Training
    Phase
    ● ଟลଌྔ๏ ● ϑΟϯΨʔϓϦϯτ๏
    - Ґஔͷ෼͔͍ͬͯΔج४ہ
    ͔Βͷڑ཭Λਪఆ
    - ෳ਺ج४ہ͔Βͷڑ཭ͰҐ
    ஔਪఆ

    View full-size slide

  19. ηϯαϊʔυͷઃஔҐஔਪఆ
    ■ WiFi APͷ৴߸Λ࢖ͬͯηϯαΛଌҐ [1,2]
    ● ηϯαΛஔ͍͚ͨͩͰͲ͜ʹ͋Δ͔͕෼͔Δ
    ※ ηϯαϊʔυ͸ZigBee
    㱺 WiFi৴߸Λड৴Ͱ͖ͳ͍
    13
    RSS1=–40
    RSS2=–50
    RSS3=–45
    RSS1=–60
    RSS2=–54
    RSS3=–42
    AP 2
    AP 1
    AP 3
    Sensor
    node
    Localiza;on
    Server
    WiFi
    Fingerprint
    DB
    ZigLoc Demo
    [1] S. Ishida et al., “WiFi AP-RSS Monitoring using Sensor Nodes toward Anchor-Free Sensor Localiza;on”, IEEE VTC-Fall, Sep
    2015.
    [2] T. Yamamoto et al., “Accuracy improvement in sensor localiza;on system u;lizing heterogeneous wireless technologies”,
    ICMU, Oct 2017.

    View full-size slide

  20. ଌҐٕज़ͷԠ༻ྫ (1)
    ■ BLEʢBluetooth Low Energyʣ
    ● iBeaconͳͲɼଌҐਫ਼౓͕ѱ͍
    ● Separate Channel Fingerprin;ng
    • νϟωϧ෼཭BLEଌҐ [3]
    14
    Receiver
    MacBook Pro
    BLE Beacon
    BLED112
    +Mobile ba7ery
    [3] S. Ishida et al., “Proposal of Separate Channel Fingerprin;ng Using Bluetooth Low Energy”, IIAI AAI, Jul 2016.
    Beacon1 RSS
    Beacon2 RSS
    Beacon3 RSS
    Fingerprints in DB
    Fingerprint x at
    unknown
    location
    Find nearest
    neighbor
    k
    Beacon1: –52dBm@ch37
    Beacon1: –62dBm@ch38
    Beacon1: –60dBm@ch39
    Beacon2: –61dBm@ch37
    Beacon2: –70dBm@ch38
    Beacon2: –58dBm@ch39
    38
    Beacon1 Beacon2
    Channel gain
    Frequency
    Frequency
    ch37 39

    View full-size slide

  21. ଌҐٕज़ͷԠ༻ྫ (2)
    ■ ΦϯσϚϯυҐஔ৘ใαʔϏε
    [4]
    15
    RSS=–48
    RSS=–52
    RSS=–58
    RSS=–55
    WiFi AP
    Core AP
    WiFi Device
    Localization Server
    On-demand Loc Demo
    [4] S. Ishida et al., “On-Demand Indoor Loca;on-based Service using Ad-Hoc Wireless Posi;oning Network”, IEEE ICESS, Aug 2015.

    View full-size slide

  22. ଌҐܭࢉͷ෼ࢄ
    ■ ΦϯσϚϯυҐஔ৘ใγεςϜ [5]
    ● Ϣʔβ͕૿Ճ͢Δͱॏ͘ͳΔ
    㱺 BBϧʔλΛ࢖ͬͯ௨৴Ͱ෼ࢄ
    16
    Router
    WNDR4300
    WiFi Mesh Node
    PCWL-0100
    172.17.0.0/16 192.168.0.0/16
    10.0.0.1
    10.0.0.2
    Router
    Localiza-on Servers
    WiFi APs
    Shuffle
    Map
    Reduce
    192.168.0.1
    192.168.0.2
    dest 10.0.0.0/9 DNAT to 192.168.0.1
    dest 10.128.0.0/9 DNAT to 192.168.0.2
    NAT Rule
    WiFi APs
    Localiza.on
    Servers
    10.0.0.0
    :
    10.127.0.0
    10.128.0.0
    :
    10.255.255.255
    [5] J. Kajimura, “Design of distributed calcula;on scheme using network address transla;on for ad-hoc wireless posi;oning
    network”, Springer CCIS, vol.760, ISIP Post Proc., Oct 2017.

    View full-size slide

  23. ं྆ηϯγϯά
    Vehicle Sensing
    17

    View full-size slide

  24. എܠ
    ■ ITS (Intelligent Transporta;on System)
    ● ಓ࿏্ͷं྆ͷݕग़͸ॏཁͳج൫ٕज़
    ͷ1ͭ
    ■ طଘͷं྆ݕग़ηϯα
    ● ϧʔϓίΠϧɼޫిηϯαɼ௒Ի೾ɼ
    ੺֎ઢͳͲ
    㱺 ઃஔɾ؅ཧʹಓ࿏޻ࣄ͕ඞཁͰߴίετ
    㱺 ݕग़ൣғ͕ڱ͍ͨΊೋྠंͷݕग़͕ࠔ೉
    㱺 ௿ίετ͔ͭं྆λΠϓʹΑΒͣߴਫ਼౓ʹं྆Λ
    ݕग़͢ΔγεςϜ͕ٻΊΒΕ͍ͯΔ
    18

    View full-size slide

  25. Իڹं྆ηϯγϯά
    ■ εςϨΦϚΠΫͰं྆ݕग़ [6]
    ● ं྆૸ߦԻͷ౸ୡ࣌ؒࠩΛඳ
    ͍ͨʮα΢ϯυϚοϓʯΛར༻
    • ૸ߦं྆͸SࣈΧʔϒΛඳ͘
    㱺 SࣈΧʔϒΛݕग़͢Δ
    19
    Microphones
    Recorder
    2-lane Road
    α΢ϯυϚοϓ
    [6] S. Ishida et al., “SAVeD: Acous;c Vehicle Detector with Speed Es;ma;on capable of Sequen;al Vehicle Detec;on”, IEEE
    ITSC, Nov 2018.
    Δtmax t
    Δt
    –Δtmax

    View full-size slide

  26. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R

    View full-size slide

  27. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R

    View full-size slide

  28. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R

    View full-size slide

  29. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R

    View full-size slide

  30. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R

    View full-size slide

  31. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R

    View full-size slide

  32. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R

    View full-size slide

  33. α΢ϯυϚοϓ
    ■ 2୆ͷϚΠΫͷं྆Իͷ࣌ؒࠩ
    ● ࣌ؒࠩͷมԽ

    = α΢ϯυϚοϓ
    Time t
    t
    Sound Delay
    20
    L
    R
    L R
    t
    =
    1
    c
    8
    <
    :
    s✓
    x
    + D
    2
    ◆2
    +
    L
    2
    s✓
    x
    D
    2
    ◆2
    +
    L
    2
    9
    =
    ;

    View full-size slide

  34. α΢ϯυϚοϓͷྫ
    ■ ϚΠΫؒͷڑ཭ D = 50 [cm]
    ● ౸ୡ࣌ؒࠩͷ࠷େ஋͸໿1.47 [ms]
    21
    -1.5
    -1
    -0.5
    0
    0.5
    1
    1.5
    30 35 40 45 50
    Sound delay ∆t [ms]
    Time t [s]

    View full-size slide

  35. α΢ϯυϚοϓͷྫ
    ■ ϚΠΫؒͷڑ཭ D = 50 [cm]
    ● ౸ୡ࣌ؒࠩͷ࠷େ஋͸໿1.47 [ms]
    21
    -1.5
    -1
    -0.5
    0
    0.5
    1
    1.5
    30 35 40 45 50
    Sound delay ∆t [ms]
    Time t [s]
    ࠨ͔Βӈ ӈ͔Βࠨ

    View full-size slide

  36. α΢ϯυϚοϓͷྫ
    ■ ϚΠΫؒͷڑ཭ D = 50 [cm]
    ● ౸ୡ࣌ؒࠩͷ࠷େ஋͸໿1.47 [ms]
    21
    -1.5
    -1
    -0.5
    0
    0.5
    1
    1.5
    30 35 40 45 50
    Sound delay ∆t [ms]
    Time t [s]
    ࠨ͔Βӈ ӈ͔Βࠨ
    RANSACͰSࣈΧʔϒΛݕग़
    㱺 Ϟσϧ͔ࣜΒ଎౓΋ਪఆ
    t
    =
    1
    c
    8
    <
    :
    s✓
    x
    + D
    2
    ◆2
    +
    L
    2
    s✓
    x
    D
    2
    ◆2
    +
    L
    2
    9
    =
    ;

    View full-size slide

  37. α΢ϯυϚοϐϯά σϞ
    22
    Real Sound Mapping Demo

    View full-size slide

  38. Ԡ༻
    ■ লిྗԽ
    ● Waveletม׵Λ࢖ͬͨলిྗं྆ݕग़ख๏ [7]
    ■ ϊΠζରࡦ
    ● ఆৗϊΠζՃࢉʹΑΔϊΠζ෼཭ [8]
    ● ෩ϊΠζରࡦ [9]
    ■ ྻंݕग़
    ● ྻं૸ߦԻΛػցֶशͯ͠ݕग़ [10]
    ● ྻं৐ंҐஔਪఆ [11]
    23
    [7] ٱอ ଞ, “཭ࢄ΢ΣʔϒϨοτม׵Λ༻͍ͨলϦιʔεं྆ݕग़γεςϜͷઃܭͱධՁ”, IPSJ ITSݚڀձ, Feb-Mar,
    2019. (will appear)
    [8] ཥ ଞ, “ϚΠΫΛ༻͍ͨं྆ݕग़γεςϜʹ͓͚Δ؀ڥϊΠζ࡟ݮख๏ͷఏҊ”, IEICE ASNݚڀձ, Jan 2019.
    [9] M. Uchino et al., “Ini;al design of acous;c vehicle detector with wind noise suppressor”, PerVehicle, Mar 2019 (will
    appear)
    [10] K. Sato et al., “Proposal of acoustic train detection system for crowdsensing”, ITS-AP Fukuoka Post Proc, 2019 (will appear)
    [11] ࠤ౻ ଞ, “ϚΠΫϩϑΥϯΛ༻͍ͨమಓ৐ंҐஔਪఆख๏ͷઃܭͱධՁ”, IPSJ ITSݚڀձ, Feb-Mar 2019 (will appear)

    View full-size slide

  39. ԻڹΞϓϩʔνͷݶք
    ■ ࣭ٙʹͯ
    ● ʮࣗసं͸ݕग़Ͱ͖ͳ͍ͷʁʯ
    ● ʮาߦऀ͸ݕग़Ͱ͖ͳ͍ͷʁʯ
    24

    View full-size slide

  40. ԻڹΞϓϩʔνͷݶք
    ■ ࣭ٙʹͯ
    ● ʮࣗసं͸ݕग़Ͱ͖ͳ͍ͷʁʯ
    ● ʮาߦऀ͸ݕग़Ͱ͖ͳ͍ͷʁʯ
    ■ ͜ͷΑ͏ͳ࣭໰Λ͞ΕΔ͕ɼݪཧతʹࠔ೉
    ● ͦ΋ͦ΋ൃͤΒΕΔԻ͕খ͍͞
    ● ଎౓͕஗͍ͷͰα΢ϯυϚοϓ্Ͱͷղੳ͕ࠔ೉
    24

    View full-size slide

  41. طଘͷंɼࣗసंɼาߦऀݕग़
    ■ ं
    ● ϧʔϓίΠϧɼϨʔβɼ௒Ի೾
    ■ ࣗసंɾาߦऀ [12,13]
    ● Χϝϥ΍Ϩʔβڑ཭ܭ (LiDAR)ʹΑΔݕग़
    ■ ઃஔʹ͸ಓ࿏޻ࣄ͕ඞཁ
    ● ಋೖɾ؅ཧίετ͕ߴ͍
    㱺 ௿ίετͳηϯα͕ٻΊΒΕ͍ͯΔ
    25
    [12] F. García et al., “Context aided pedestrian detec;on for danger es;ma;on based on laser scanner and computer vision”,
    Expert Systems with Applica;ons, Nov 2014.
    [13] P. Dollár et al., “Pedestrian detec;on: an evalua;on of the state of the art”, IEEE Trans. Parern Analysis and Machine
    Intelligence, Apr 2012.

    View full-size slide

  42. ■ WiFi௨৴ͷ఻ൖ࿏ͷมԽʹΑΔηϯγϯά
    ● ंɼࣗసंɼਓͰճંɾ൓ࣹ͕ൃੜ
    ৽͍͠Ξϓϩʔν
    26

    View full-size slide

  43. ■ WiFi௨৴ͷ఻ൖ࿏ͷมԽʹΑΔηϯγϯά
    ● ंɼࣗసंɼਓͰճંɾ൓ࣹ͕ൃੜ
    ճંɾ൓ࣹͯ͠౸ୡ
    ৽͍͠Ξϓϩʔν
    26

    View full-size slide

  44. ■ WiFi௨৴ͷ఻ൖ࿏ͷมԽʹΑΔηϯγϯά
    ● ंɼࣗసंɼਓͰճંɾ൓ࣹ͕ൃੜ
    ճંɾ൓ࣹͯ͠౸ୡ
    ৽͍͠Ξϓϩʔν
    26
    ো֐෺ͷҐஔ΍େ͖͞
    ʹΑͬͯճંɼ൓ࣹঢ়
    ଶ͸มԽ

    View full-size slide

  45. ؔ࿈ݚڀ: ԰಺ແઢηϯγϯά
    ■ E-eyes [14]
    ● ԰಺ʹ͓͍ͯ఻ൖ࿏ͷมԽͰਓؒͷߦಈΛਪఆ
    ■ Smokey [15]
    ● NLOS (Non Line-of-Sight)؀ڥͰ٤Ԏಈ࡞Λݕग़
    ■ RF-Pose [16]
    ● นӽ͠ʹෳ਺ਓͷ࢟੎Λਪఆ (ઐ༻ແઢػΛ࢖༻)
    ■ SignFi [17]
    ● WiFiͷ఻ൖ࿏มԽͰख࿩Λೝࣝ
    27
    [14] Y. Wang et al., “E-eyes: In-home device-free ac;vity iden;fica;on using fine-grained WiFi signatures”, ACM MobiCom, Sep
    2014.
    [15] X. Zheng et al., “Smokey: Ubiquitous smoking detec;on with commercial WiFi infrastructures”, IEEE INFOCOM, Jul 2016.
    [16] M. Zhao et al., “Through-wall human pose es;ma;on using radio signals”, CVPR, Jun 2018.
    [17] Y. Ma et al., “SignFi: sign language recogni;on using WiFi”, ACM IMWUT, Mar 2018.

    View full-size slide

  46. ■ ACT-Iʮ৘ใͱະདྷʯʹ࠾୒ʢ2018೥10݄ʙʣ
    ● ແઢ௨৴Λ༻͍ͨं྆ɾࣗసंɾาߦऀݕग़ٕज़
    ● ʮʙʢલུʣʙຊݚڀͷ໨త͸ɺࢢൢ
    ͷແઢ௨৴ػثΛվ଄͢Δ͜ͱͳ͘༻
    ͍ͯं྆΍ࣗసंɺาߦऀΛݕग़͢Δ
    γεςϜΛ։ൃ͢Δ͜ͱͰ͢ɻແઢ௨
    ৴͸पғͷ؀ڥมԽͷӨڹΛड͚Δ͜
    ͱ͔Βɺແઢ௨৴͕ड͚ͨӨڹΛղੳ
    ͢Δ͜ͱͰं྆ɺࣗసंɺาߦऀͷݕ
    ग़Λ࣮ݱ͠·͢ɻʯ
    㱺 ݚڀΛ։࢝ͨ͠͹͔ΓͷͨΊ੒Ռͷ͘͘͝͝Ұ෦Λ
    ঺հ
    ઓུత૑଄ݚڀਪਐࣄۀ
    28
    ग़ల: ACT-I 平成30年度採択課題

    View full-size slide

  47. WiFiΛ༻͍ͨं྆ݕग़ [18]
    ■ ۝େ಺ͷಓ࿏Ͱ࣮ݧ
    ● ಓ࿏ͷ྆ଆʹૹड৴ػΛઃஔ
    ● ఻ൖ࿏৘ใΛղੳͯ͠௨աं྆Λݕग़
    29
    TX
    RX
    [18] M. Cong et al., “Proposal of On-road Vehicle Detection Method Using WiFi Signal”, IPSJ ITSݚڀձ, Feb-Mar 2019 (will appear)

    View full-size slide

  48. WiFiʹΑΔं྆ݕग़ͷ֓ཁ
    1. WiFi (802.11n)୺຤ؒͰ50msִؒͰ௨৴
    2. ఻ൖ࿏৘ใͷҐ૬৘ใΛநग़
    3. ิਖ਼΍ϊΠζআڈͳͲ্ͨ͠Ͱಛ௃ྔΛநग़
    ● Window಺ͷ࠷େ஋ɼ࠷খ஋ɼ෼ࢄͳͲ
    4. ػցֶश
    㱺 ं྆छผ͝ͱʹݕग़Ͱ͖Δ͔Λࢼߦ
    ● େܕंɼόϯɼී௨ंɼখܕंͷ4छผ
    30
    Phase
    Extraction
    Phase
    Calibration
    PCA
    Feature
    Extraction
    Machine
    Learning
    Windowing

    View full-size slide

  49. ■ ं྆छผ͝ͱͷݕग़ (ͨͩ͠unbalanced data)
    ● Precision = 54.4%, Recall = 48.5%, Accuracy = 55.3%
    㱺 ୯७ͳݕग़ํ๏Ͱ͸શવμϝ (Work in Progress)
    ं྆ݕग़݁Ռ
    31
    EsHmated
    Actual
    େܕं όϯ ී௨ं খܕं
    େܕं 9 0 2 0
    όϯ 2 3 5 0
    ී௨ं 1 0 14 2
    খܕं 1 0 8 0

    View full-size slide

  50. ԰֎ਓମηϯγϯά
    Outdoor Human Sensing
    32

    View full-size slide

  51. ԰֎ਓମηϯγϯά [19]
    ■ WiFi (802.11ac)Λ࢖ͬͨ԰֎ਓମηϯγϯά
    ● ηϯγϯάྖҬ಺ʹਓ͕͍Δ͔Λਪఆ
    ● ं྆ݕग़ͱ΄΅ಉ͡࢓૊Έ (ͨͩ͠802.11acΛ࢖༻)
    33
    Training Phase
    Estimation Phase
    Data Acquisition Block
    Pre-Process
    Block
    Machine
    Learning
    Block
    Machine
    Learning
    Block
    AP
    CSI
    monitoring
    station
    CSI
    measuring
    station
    Target area
    Human
    Location
    Actual
    human location
    Angles
    ij, ij
    AAAB/nicdVDLSsNAFL2pr1pf8bFzM1gKLkpJ3Oiy4MZlBfuAJoTJdNKOnUzCzESooeCnuBIUxK3/4cq/cfpQWsUDFw7n3Mu5nDDlTGnH+bQKK6tr6xvFzdLW9s7unr1/0FJJJgltkoQnshNiRTkTtKmZ5rSTSorjkNN2OLyc+O07KhVLxI0epdSPcV+wiBGsjRTYR146YEHObsdV5KVqRgO77NacKZDzh3xbZZijEdgfXi8hWUyFJhwr1XWdVPs5lpoRTsclL1M0xWSI+7RrqMAxVX4+/X6MKkbpoSiRZoRGU3XxIsexUqM4NJsx1gP125uIP15lKUpHF37ORJppKsgsKco40gmadIF6TFKi+cgQTCQzzyIywBITbRorLbbwP2md1Vyn5l475Xp13kcRjuEETsGFc6jDFTSgCQTu4RGe4cV6sJ6sV+tttlqw5jeHsATr/Qsol5Vq
    AAAB/nicdVDLSsNAFL2pr1pf8bFzM1gKLkpJ3Oiy4MZlBfuAJoTJdNKOnUzCzESooeCnuBIUxK3/4cq/cfpQWsUDFw7n3Mu5nDDlTGnH+bQKK6tr6xvFzdLW9s7unr1/0FJJJgltkoQnshNiRTkTtKmZ5rSTSorjkNN2OLyc+O07KhVLxI0epdSPcV+wiBGsjRTYR146YEHObsdV5KVqRgO77NacKZDzh3xbZZijEdgfXi8hWUyFJhwr1XWdVPs5lpoRTsclL1M0xWSI+7RrqMAxVX4+/X6MKkbpoSiRZoRGU3XxIsexUqM4NJsx1gP125uIP15lKUpHF37ORJppKsgsKco40gmadIF6TFKi+cgQTCQzzyIywBITbRorLbbwP2md1Vyn5l475Xp13kcRjuEETsGFc6jDFTSgCQTu4RGe4cV6sJ6sV+tttlqw5jeHsATr/Qsol5Vq
    AAAB/nicdVDLSsNAFL2pr1pf8bFzM1gKLkpJ3Oiy4MZlBfuAJoTJdNKOnUzCzESooeCnuBIUxK3/4cq/cfpQWsUDFw7n3Mu5nDDlTGnH+bQKK6tr6xvFzdLW9s7unr1/0FJJJgltkoQnshNiRTkTtKmZ5rSTSorjkNN2OLyc+O07KhVLxI0epdSPcV+wiBGsjRTYR146YEHObsdV5KVqRgO77NacKZDzh3xbZZijEdgfXi8hWUyFJhwr1XWdVPs5lpoRTsclL1M0xWSI+7RrqMAxVX4+/X6MKkbpoSiRZoRGU3XxIsexUqM4NJsx1gP125uIP15lKUpHF37ORJppKsgsKco40gmadIF6TFKi+cgQTCQzzyIywBITbRorLbbwP2md1Vyn5l475Xp13kcRjuEETsGFc6jDFTSgCQTu4RGe4cV6sJ6sV+tttlqw5jeHsATr/Qsol5Vq
    AAAB/nicdVDLSsNAFL2pr1pf8bFzM1gKLkpJ3Oiy4MZlBfuAJoTJdNKOnUzCzESooeCnuBIUxK3/4cq/cfpQWsUDFw7n3Mu5nDDlTGnH+bQKK6tr6xvFzdLW9s7unr1/0FJJJgltkoQnshNiRTkTtKmZ5rSTSorjkNN2OLyc+O07KhVLxI0epdSPcV+wiBGsjRTYR146YEHObsdV5KVqRgO77NacKZDzh3xbZZijEdgfXi8hWUyFJhwr1XWdVPs5lpoRTsclL1M0xWSI+7RrqMAxVX4+/X6MKkbpoSiRZoRGU3XxIsexUqM4NJsx1gP125uIP15lKUpHF37ORJppKsgsKco40gmadIF6TFKi+cgQTCQzzyIywBITbRorLbbwP2md1Vyn5l475Xp13kcRjuEETsGFc6jDFTSgCQTu4RGe4cV6sJ6sV+tttlqw5jeHsATr/Qsol5Vq
    [19] M. Miyazaki et al., “Ini;al arempt on outdoor human detec;on using IEEE 802.11ac WLAN signal”, IEEE SAS, Mar 2019.
    (will appear)

    View full-size slide

  52. ࣮ݧ؀ڥ
    ■ 9ͭͷΤϦΞ + ਓͳ͠ (ϥϕϧ0)Λ

    ଟΫϥε෼ྨ໰୊ͱͯ͠ػցֶश
    34
    CSI measuring station
    CSI monitoring station
    AP
    4
    1 7
    5
    2 8
    6
    3 9
    30 meters
    30 meters
    6 meters
    6 meters
    0 : no human
    STA1 STA2
    ΤϦΞ
    ൪߸

    View full-size slide

  53. ■ ਫ਼౓ (ਖ਼͍͠ݕग़ͷׂ߹)
    ● 99.86% (STA2ͷΈ࢖ͬͨ৔߹)
    ● 56.01% (STA1ͷ
    Έ࢖ͬͨ৔߹)
    ݕग़݁Ռ
    35
    0 1 2 3 4 5 6 7 8 9
    Estimated Area
    0
    1
    2
    3
    4
    5
    6
    7
    8
    9
    Actual Area
    547623 50 1 11 15 0 0 0 0 0
    7 551917 27 0 330 0 19 0 0 0
    0 69 550299 368 1552 4 0 8 0 0
    3 0 520 546006 18 297 42 114 0 0
    0 30 1711 36 550511 0 112 0 0 0
    3 0 6 566 0 547076 0 577 47 25
    3 0 5 45 221 0 546626 0 0 0
    0 0 1 92 0 703 4 546500 0 0
    0 0 0 17 0 1 5 11 546966 0
    0 0 0 0 0 5 0 0 2 553393

    View full-size slide

  54. ·ͱΊ
    ■ IoT࣮ݱʹ޲͚ͯ͸ηϯγϯά΋େ੾
    ■ ࣮ۭؒ৘ใΛऔಘ͢ΔແઢηϯαωοτϫʔΫ
    ● Ͳ͜ʹஔ͍ͨͷ͔஌Βͳ͍ͱ͍͚ͳ͍
    ● ηϯγϯάର৅Λ֦େ͍ͨ͠
    ■ ԰಺ଌҐٕज़
    ● ҟछແઢؒ௨৴Λ༻͍ͨηϯαଌҐ
    ● νϟωϧؒಛੑࠩΛར༻ͨ͠BLEଌҐ
    ■ ৽͍͠ηϯγϯάٕज़
    ● ं྆ηϯγϯά (Ի, ແઢ [Work in Progress])
    ● ਓମηϯγϯά (ແઢ)
    37

    View full-size slide

  55. Thank you!!
    QuesHons & Answers
    38

    View full-size slide

  56. © 2019 Shigemi ISHIDA licensed under CC BY-NC 4.0

    View full-size slide