Slide 1

Slide 1 text

IEICE CQ IoT CSI Sept 24, 2024

Slide 2

Slide 2 text

n p 2024 9 @ p 2024 9 24 p September 24, 2024 2

Slide 3

Slide 3 text

IoT September 24, 2024 12

Slide 4

Slide 4 text

September 24, 2024 13

Slide 5

Slide 5 text

n p n IoT p IoT (?) September 24, 2024 14 Livingroom Bedroom Bath- room Wash- room Dining/Kitchen Nature Remo Switchbot Switchbot Hub Echo Flex Echo Flex Smart power strip (?) (?)

Slide 6

Slide 6 text

IoT (PnP IoT) n IoT September 24, 2024 15 TV 4 ( ) ( OFF)

Slide 7

Slide 7 text

September 24, 2024 16 WPS OK ( ) ( OFF) تُ٭عتم٭؜ ԛ ԛ ԛ

Slide 8

Slide 8 text

September 24, 2024 17 WPS OK ( ) ( OFF) تُ٭عتم٭؜ ԛ ԛ ԛ

Slide 9

Slide 9 text

September 24, 2024 18

Slide 10

Slide 10 text

n p ( ) n ( ) p September 24, 2024 19

Slide 11

Slide 11 text

[zhang 19] n p n p l l n p F ( ) ≃ 89.5% September 24, 2024 20 [zhang 19] Danger-pose detection system using commodity Wi-Fi for bathroom monitoring, Sensors, 19(4) https://doi.org/10.3390/s19040884 CSI Illustration by Storyset

Slide 12

Slide 12 text

Figure 8. Experiment setup: (a) transmitter (TX), receiver (RX), and server used in our experiment, and (b) configuration of TX, RX, and bathtub. For the dangerous situation simulation, we assumed three dangerous situations as shown in Figure 9: (1) keep the lying position in a long time, (2) sink the whole body below the water surface, and (3) sink the face below the water surface. The danger-pose detection system is used to detect dangerous situations when taking a bath. Although target dangerous situations are not limited to the three simulated situations, we conducted evaluations with these three situations to demonstrate the basic performance of our system as an initial evaluation. (a) (b) (c) Figure 9. Simulated three dangerous situations: (a) steady lying position, (b) the whole body sinks below the water surface, and (c) the face sinks below the water surface. We collected CSI data at a rate of 20 Hz on channel 40 in a 5-GHz band for couple of hours on six different days during a three-month period. During the three-month period, environmental changes including location of furniture and daily objects have been occurred. Locations of transmitter and receiver might also include errors up to approximately 10 centimeters as we put transmitter and receiver on each of the six days. For not in bath, safe, and danger activities, we collected 119,324, 288,291, and 21,572 CSI data samples, respectively, in total in six days. [zhang 19] Danger-pose detection system using commodity Wi-Fi for bathroom monitoring, Sensors, 19(4) https://doi.org/10.3390/s19040884

Slide 13

Slide 13 text

[miyazaki 18][ishida 21] n p n p p p September 24, 2024 22 [miyazaki'18] Initial Attempt on Outdoor Human Detection using IEEE 802.11ac WLAN Signal, IEEE SAS [ishida 21] IEEE 802.11ac-based outdoor device-free human localization, Sensors and Materials, 33(1) CSI Illustration by Storyset

Slide 14

Slide 14 text

n p CSI : 4 p 36m2×9 p n p CSI : 4 p 30m2×11 September 24, 2024 23 WLAN AP CSI Measuring Stations CSI Monitoring Station 1 4 7 2 5 8 3 6 9 CSI Measuring Station STA1 WLAN AP CSI Monitoring Station 30m 6m 6m Label 0: no human 30m CSI Measuring Station STA3 CSI Measuring Station STA2 4.6m 4.6m 5.9m 7.0m 6.9m 6.9m 7.0m 2.3m 1 2 3 4 5 6 7 8 9 10 11 STA2 CSI Measuring Station STA1 WLAN AP CSI Monitoring Station Label 0: no human STA4 STA3 Wall Pillar

Slide 15

Slide 15 text

[uchino 20] n p p n 2 p F : 0.83 September 24, 2024 24 [uchino 20] Initial Design of Two-Stage Acoustic Vehicle Detection System for High Traffic Roads, PerVehicle, IEEE PerCom Workshop D/2 M1 M2 x O d2 d1 D/2 L ౸དྷํ޲ͷมԽΛఆࣜԽ Time t Sound delay Δt ϩόετਪఆʹΑΔ ϑΟοςΟϯά Sound Mapper Vehicle Detector M 2 M 1 Sound Retriever LPFs Out Sound Map 2441 2442 2443 2444 2445 2446 Time [s] −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 Sound Delay [ms] ֶ ੜ F஋0.83Ͱਐߦํ޲Λࣝผ͠ͳ͕Βं྆Λݕग़

Slide 16

Slide 16 text

n 4 n 60 n ⾒ 609 n p めっちゃ時間 かかりました! September 24, 2024 25 [uchino 20] Initial Design of Two-Stage Acoustic Vehicle Detection System for High Traffic Roads, PerVehicle, IEEE PerCom Workshop TP 133 132 265 FN 165 179 344 FP 81 80 161 Precision 0.62 0.62 0.62 Recall 0.45 0.42 0.44 F-measure 0.52 0.50 0.51 harmonic mean of precision and recall, which provides a comprehensive evaluation of the classifier. B. Detection Performance Table I shows the system performance, i.e., the number of TP, FN, and FP detections as well as the calculated precision, recall, and F-measure for ✓t = 1.5s. • The precision of the proposed Two-Stage Acoustic Ve- hicle Detection System is 0.76, 14 points higher than SAVeD. By resetting the detection window in the Post- Fitting block and detecting vehicles using only sound map points in the neighborhood of the estimated vehicle passing time, the amount of FP detections was reduced. • The recall of 0.53 is a 9-point improvement compared to previous work. This is due to the system’s ability to make use of all the sound map points corresponding to a given vehicle. • The F-measure of 0.63 is 12 points higher than SAVeD. The relatively low F-measures exhibited by both the Two- Stage Acoustic Vehicle Detection System and SAVeD are due to their low recall values compared to their precision values. As the number of simultaneously passing vehicles increases, the number of sound map points corresponding to a single vehicle decreases because only one point is drawn on the sound map at each time step. As a result, the S-curve becomes sparse, increasing the probability of a FN detection. • The Two-Stage Acoustic Vehicle Detection System de- Fig. 10. Proposed system F-measure as a function of passing time error margin ✓t Microphone setup location Fig. 11. System monitoring viewpoint The above results show that our Two-Stage Acoustic Vehicle Detection System is an improvement over the SAVeD in terms of vehicle detection performance. C. Passing Time Error The detection performance of our system depends strongly on the estimated passing time used in the Post-Fitting block. As described in Section IV-A, the smaller the passing time error margin ✓ts, the higher the number of FP detections. It is therefore important to examine the influence of the passing time error margin on detection performance. Figure 10 shows the F-measures for ✓t between 0.5 s and 2.0 s. The F-measure starts to increase at about ✓t = 0.6 s and stabilizes around ✓t = 1.2 s. The optimum value of ✓t corresponds to the point at which the F-measure just begins to stabilize, as a too large value of ✓t will cause the system

Slide 17

Slide 17 text

... n IoT September 24, 2024 26

Slide 18

Slide 18 text

CSI IoT [ishida 22] Room-by-Room Device Grouping for Put-and- Play IoT System, IEEE Globecom 27 September 24, 2024

Slide 19

Slide 19 text

n WiFi (CSI) p èCSI (= )IoT 28 E F v v A B C D F E G A D E G B C F September 24, 2024

Slide 20

Slide 20 text

n WiFi (IEEE 802.11n, 802.11ac) p ( ) p p 29 CSI (Channel State Information) E[k] R[k] E[k] R[k] ࡶԻn H ௨৴࿏ʢνϟωϧʣ H n Hl = 2 6 6 4 h11 h12 · · · h1j h21 h22 · · · h2j · · · · · · · · · · · · hi1 hi2 · · · hij 3 7 7 5 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 RXΞϯςφ਺ TXΞϯςφ਺ 𝑅 𝑘 = 𝐻! 𝐸 𝑘 + 𝑛 𝑙: September 24, 2024

Slide 21

Slide 21 text

n CSI p PinLoc [sen 12], PhaseFi [wang 16], DeepFi [wang 17] ➔ n p FUSIC [jiokeng 20], SpotFi [kotaru 15] ➔ September 24, 2024 30 [sen 12] You are facing the Mona Lisa: Spot localization using PHY layer information, ACM MobiSys [wang 16] CSI phase fingerprinnting for indoor localization with a deep learning approach, IEEE Internet Things J., 3(6) [wang 17] CSI-based fingerprinting for indoor localization: A deep learrning approach, IEEE Trans. Veh. Technol., 66(1) [jiokeng 20] When FTM discovered MUSIC: Accurate WiFi-based ranging in the presence of multipath, IEEE INFOCOM [kotaru 15] SpotFi: decimeter level localization using WiFi, ACM SIGCOMM CC Review

Slide 22

Slide 22 text

31 September 24, 2024

Slide 23

Slide 23 text

September 24, 2024 32 sin ∅"# , cos ∅"# instead of ∅"# mean median max min std p2p iqr

Slide 24

Slide 24 text

September 24, 2024 33 k-means

Slide 25

Slide 25 text

n 1LDK p AP CSI dining p dining, bedroom, living Galaxy S7 edge l : 0 90cm p 10Hz 5 CSI p 34 September 24, 2024

Slide 26

Slide 26 text

n * p NH/* = NH/OP + NH/CL p */OP = NH/OP + DN/OP + LV/ OP + BD/OP September 24, 2024 35 Dataset (5 min each) Abbrv Human walking in Doors No human w/ opened doors NH/OP Opened No human w/ closed doors NH/CL Closed Dining room w/ opened doors DN/OP Dining room Opened Dining room w/ closed doors DN/CL Dining room Closed Living room w/ opened doors LV/OP Living room Opened Living room w/ closed doors LV/CL Living room Closed Bedroom w/ opened doors BD/OP Bedroom Opened Bedroom w/ closed doors BD/CL Bedroom Closed

Slide 27

Slide 27 text

n (ARI) p p −1 ≤ ARI ≤ 1 l 1 l 0 n p September 24, 2024 36 Living room Bedroom

Slide 28

Slide 28 text

1: n Win size = 10s, 𝑁!"# = 4 n IoT n 100 ARI September 24, 2024 37 Feature */OP */CL */* sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 mean 0.10 0.40 0.44 0.01 0.44 0.35 median 0.10 0.44 0.44 0.03 0.45 0.36 max 0.47 0.39 0.41 0.43 0.44 0.39 min 0.60 0.43 0.39 0.35 0.52 0.44 std 0.69 0.89 0.34 0.81 0.63 0.93 p2p 0.69 0.83 0.52 0.76 0.81 0.89 iqr 0.45 0.85 0.28 0.87 0.36 0.93

Slide 29

Slide 29 text

1: n Win size = 10s, 𝑁!"# = 4 n IoT n 100 ARI September 24, 2024 38 Feature */OP */CL */* sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 sin ∅!" , cos ∅!" 𝝍𝒍𝒋 mean 0.10 0.40 0.44 0.01 0.44 0.35 median 0.10 0.44 0.44 0.03 0.45 0.36 max 0.47 0.39 0.41 0.43 0.44 0.39 min 0.60 0.43 0.39 0.35 0.52 0.44 std 0.69 0.89 0.34 0.81 0.63 0.93 p2p 0.69 0.83 0.52 0.76 0.81 0.89 iqr 0.45 0.85 0.28 0.87 0.36 0.93 CSI

Slide 30

Slide 30 text

2: n IoT n 100 ARI 39 Human location ARI 0.42 Dining room 0.27 Living room 1.00 Bedroom 0.29 Anywhere 0.95 Anywhere or no where 0.95 September 24, 2024

Slide 31

Slide 31 text

( ) n p CSI p n p CSI = p CSI September 24, 2024 40

Slide 32

Slide 32 text

[ 23] IoT IoT CSI , IEICE CS 41 September 24, 2024

Slide 33

Slide 33 text

September 24, 2024 42 1) 0.2 0.9 … 0.1 0.8 … … … … … … 0.2 0.9 … 0.1 0.8 … … … … … … AAAB8XicbVDLSgNBEOyNr7i+oh69DIaAp7Ab4uMY8OIxgnlIsoTZyWwyZGZ2mZkVwpKv8CQoiFc/x5N/4yTZgyYWNBRV3XR3hQln2njet1PY2Nza3inuunv7B4dHpeOTto5TRWiLxDxW3RBrypmkLcMMp91EUSxCTjvh5Hbud56o0iyWD2aa0EDgkWQRI9hY6bGfaDbIav5sUCp7VW8BtE78nJQhR3NQ+uoPY5IKKg3hWOue7yUmyLAyjHA6cyv9VNMEkwke0Z6lEguqg2xx8QxVrDJEUaxsSYMWqvtrIsNC66kIbafAZqxXvbn4n9dLTXQTZEwmqaGSLBdFKUcmRvP30ZApSgyfWoKJYvZYRMZYYWJsSK5NwV/9eZ20a1X/qnp5Xy836nkeRTiDc7gAH66hAXfQhBYQEPAMr/DmaOfFeXc+lq0FJ585hT9wPn8A7eqQXQ== 21 Subcarrier –28 std p2p iqr … Subcarrier 28 std p2p iqr AAAB8XicbVDLSgNBEOyNr7i+oh69DIaAp7Ar8XEMePEYwTwkWcLsZDYZMzO7zMwKYclXeBIUxKuf48m/cZLsQRMLGoqqbrq7woQzbTzv2ymsrW9sbhW33Z3dvf2D0uFRS8epIrRJYh6rTog15UzSpmGG006iKBYhp+1wfDPz209UaRbLezNJaCDwULKIEWys9NBLNOtn/HHaL5W9qjcHWiV+TsqQo9EvffUGMUkFlYZwrHXX9xITZFgZRjidupVeqmmCyRgPaddSiQXVQTa/eIoqVhmgKFa2pEFz1f01kWGh9USEtlNgM9LL3kz8z+umJroOMiaT1FBJFouilCMTo9n7aMAUJYZPLMFEMXssIiOsMDE2JNem4C//vEpa51X/snpxVyvXa3keRTiBUzgDH66gDrfQgCYQEPAMr/DmaOfFeXc+Fq0FJ585hj9wPn8AnWWQ0A== lj … Time … Time Time Subcarrier –28 Subcarrier 28 … … AAAB8XicbVDLSgNBEOyNr7i+oh69DIaAp7Ab4uMY8OIxgnlIsoTZyWwyZGZ2mZkVwpKv8CQoiFc/x5N/4yTZgyYWNBRV3XR3hQln2njet1PY2Nza3inuunv7B4dHpeOTto5TRWiLxDxW3RBrypmkLcMMp91EUSxCTjvh5Hbud56o0iyWD2aa0EDgkWQRI9hY6bGfaDbIav5sUCp7VW8BtE78nJQhR3NQ+uoPY5IKKg3hWOue7yUmyLAyjHA6cyv9VNMEkwke0Z6lEguqg2xx8QxVrDJEUaxsSYMWqvtrIsNC66kIbafAZqxXvbn4n9dLTXQTZEwmqaGSLBdFKUcmRvP30ZApSgyfWoKJYvZYRMZYYWJsSK5NwV/9eZ20a1X/qnp5Xy836nkeRTiDc7gAH66hAXfQhBYQEPAMr/DmaOfFeXc+lq0FJ585hT9wPn8A7eqQXQ== 21 AAAB8XicbVDLSgNBEOyNr7i+oh69DIaAp7Ar8XEMePEYwTwkWcLsZDYZMzO7zMwKYclXeBIUxKuf48m/cZLsQRMLGoqqbrq7woQzbTzv2ymsrW9sbhW33Z3dvf2D0uFRS8epIrRJYh6rTog15UzSpmGG006iKBYhp+1wfDPz209UaRbLezNJaCDwULKIEWys9NBLNOtn/HHaL5W9qjcHWiV+TsqQo9EvffUGMUkFlYZwrHXX9xITZFgZRjidupVeqmmCyRgPaddSiQXVQTa/eIoqVhmgKFa2pEFz1f01kWGh9USEtlNgM9LL3kz8z+umJroOMiaT1FBJFouilCMTo9n7aMAUJYZPLMFEMXssIiOsMDE2JNem4C//vEpa51X/snpxVyvXa3keRTiBUzgDH66gDrfQgCYQEPAMr/DmaOfFeXc+Fq0FJ585hj9wPn8AnWWQ0A== lj … Feature vector 2) CSI = CSI 3) 𝑁$%&

Slide 34

Slide 34 text

n CSI p CSI n p ➔CSI ➔ CSI September 24, 2024 43 … …

Slide 35

Slide 35 text

CSI September 24, 2024 44 ... Time Device 1 Device 2 ... Device n ICA ICA i ICA j Device 1 Device 2 Device n ... ... ... ... 1) CSI (ICA) 2) ICA 3) CSI

Slide 36

Slide 36 text

: 2LDK ( ) n LAN AP 1 : Buffalo WXR-5700AX7S n IoT 9 : Raspberry Pi 3A+ ( : 0 2m) n : Intel Compute Stick n 4 (40 1 30 1 10 2 ) n CSI 24 n : 60 n 500 p ARI September 24, 2024 45 Bedroom Storeroom CL BR WC CL Living Dining Kitchen AP Data Retriever IoT Devices 2 1,3 4,6 5 7 8 9 5.3m 3.5m 2.5m 4.1m ID 1 290663 2358 2 2968 0 3 198929 2357 4 268754 2880 5 0 0 6 308506 2880 7 347484 2880 8 369627 2880 9 247124 2880

Slide 37

Slide 37 text

1. ICA ( ) 2. PCA p ICA PCA p ICA 3. p CSI ※ 𝑁$%& = 12 p >0.8 ※ September 24, 2024 46

Slide 38

Slide 38 text

: IoT n ICA 𝑁!"#$ ARI p ICA( ) 𝑁!"#$ = 1 ARI=0.943 l 𝑁1234 = 2 ARI=0.991 l 𝑁1234 ≥ 4 ARI=1.00 p PCA p Random September 24, 2024 47 0 2 4 6 8 10 Nsamp 0.7 0.8 0.9 1.0 Mean ARI ICA PCA Random 44.4%

Slide 39

Slide 39 text

n IoT IoT p CSI p n p p window p l 44.4% ARI=1.00 September 24, 2024 48

Slide 40

Slide 40 text

49 September 24, 2024

Slide 41

Slide 41 text

n IoT p IoT p IoT p n p CSI IoT n p : [email protected] p Web: https://pman0214.netlify.app/ September 24, 2024 50

Slide 42

Slide 42 text

© 2024 Shigemi ISHIDA, distributed under CC BY-NC 4.0