ICPADS 2012 FIMD: Fine-grained Device-free Motion Detection Jiang Xiao, Kaishun Wu, Youwen Yi, Lu Wang, Lionel M. Ni Department of Computer Science and Engineering Hong Kong University of Science and Technology Dec. 18 th, 2012
Outline Introduction & Motivation Related Work System Design Performance Evaluation Conclusion 2
Introduction Security for intrusion I won t be caught Device-free Motion Detection is in need! 3
Current Detection Methods Ultrasonic Camera Infrared Sensors Zhang et. al [PerCom 07, 09, 11] High cost for large environment Environment constraints (e.g., dark, smoke) Line-of-sight High cost for dense deployment Can t work well in large scale, complicate, typical indoor environment! 4
Motivation WLAN Advantages Low Cost Easy Implementation 5
Outline Introduction & Motivation Related Work System Design Performance Evaluation Conclusion 6
Related Work Received Signal Strength Indicator (RSSI) Static RF Signal Transmitter Motion Happens Receiver Youssef et. al. [MobiCom 07][PerCom 09][PerCom 12] Idea: RSS becomes anomalous when the environment changes 7
Limitations of RSS-based Techniques 1. Narrowband interference high false alarm rate Hard to distinguish narrowband interference from motion dynamic 2. RSS is of high variability Node ID miss detection Slow dynamic is easily hidden by the inherent RSS variance 8
A reliable method is in need. resist from the narrowband interference in the 2.4GHz band temporal stable in static while sensitive to a motion instantly 9
Challenge 1 Could we find such a reliable method? 10
Key Insight Receiver single value RSSI multiple values CSIs 2.4GHz S/P FFT RF band Baseband CSI Property 1 Frequency diversity CSI-based Indoor Localization: FILA[INFOCOM 12] 11
Key Insight Data in OFDM Transmitter + Channel OFDM Receiver Data out In OFDM system, the received signal over multiple subcarriers is CSI Channel gain amplitude phase 12
RSSI (dbm) CSI amplitude Key Insight Time Duration (s) RSS: variant Time Duration (s) CSI: relatively stable CSI Property 2 Temporal Stability 13
We want to harness fine-grained CSI for device-free indoor motion detection. 14
Goal To improve detection accuracy with low cost To improve detection rate To reduce false alarm rate 15
Challenges Our Contributions 1. Could we find a reliable method for device-free motion detection? Exploit the possibility of CSI 2. How to resist from narrowband interference? Extract suitable CSI features 3. How to distinguish motion event from noise? Propose burst detection alg. 16
Outline Introduction & Motivation Related Work System Design Performance Evaluation Conclusion 17
RF Signal FIMD Design DP CSI Collection CSI Feature Extraction Server Static Map Construction Burst Detection Static Map Update AP False alarm Filter Release Static Map Alarm Update 18
Outline Introduction & Motivation Related Work System Design CSI Feature Extraction Burst Detection False Alarm Filter Performance Evaluation Conclusion 19
Challenge 2 How to extract the features of CSI that resist from narrowband interference? 20
subcarriers 1. CSI Feature Extraction Target: to extract CSI features can reflect static/dynamic patterns W H 1,1, H 2,1,, H k,1, H k+1,1, H k+n,1, raw CSIs H 1,2, H 2,2,, H k2, H k+1,2, H k+n,2, H 1,30, H 2,30,, H k,30 H k+1,30, H k+n,30, Start Length n+1 Process CSIs [H k, H k+1, H k+n ] 21
1. CSI Feature Extraction Target: to extract CSI features can reflect static/dynamic patterns [H k, H k+1, H k+n ] Process CSIs Compute correlation factor btw. each column of V= (eigen (C)/n+1) Feature value vector V 22
1. CSI Feature Extraction Target: to extract CSI features can reflect static/dynamic patterns [H k, H k+1, H k+n ] V= (eigen (C)/n+1) 1st max & 2nd max eigenvalues are chosen! 23
2 nd max eigenvalue 1. CSI Feature Extraction Target: to extract CSI features can reflect static/dynamic patterns 1 st max eigenvalue CSI feature can reveal normal/motion behavior 24
1. CSI Feature Extraction In presence of narrowband interference, e.g., Bluetooth RSS: more variant CSI: relatively stable Robust to narrowband interference 25
Outline Introduction & Motivation Related Work System Design CSI Feature Extraction Burst Detection False Alarm Filter Performance Evaluation Conclusion 26
2nd eigenvalue burst 1 st eigenvalue 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 20 40 60 80 100 120 140 static dynamic static 27
2 nd max eigenvalue Feature Y 0.45 Ground Truth 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 Ground Truth 0 0.4 0.5 0.6 0.7 0.8 0.9 1 Feature X 1 st max eigenvalue 28
Challenge 3 How to distinguish motion event from noise? 29
2. Burst Detection Target: to monitor the burst motion occurrence Fact: patterns of motion are diff. from static ones Method: apply DBSCAN alg. [Ester, et al. (KDD 96)] Classify CSI Patterns No prior knowledge of the # of clusters Discovery arbitrary shape clusters 30
2. Burst Detection Target: to monitor the burst motion occurrence Method: DBSCAN clustering Density = the # of points within a specified radius ε ε max. radius of the neighborhood Nε (V i ): {V j belongs to Nε Dist(V i,v j ) ε} minpts min. # of points in an ε-neighbor to form a cluster for each point in a cluster, ε-neighbouor has to contain minpts. 31
2 nd max Feature eigenvalue Y 2. Burst Detection Target: to monitor the burst motion occurrence Euclidean distance for DBSCAN clustering 0.45 DBSCAN Result 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0.4 0.5 0.6 0.7 0.8 0.9 1 Feature X 1 st max eigenvalue 32
Outline Introduction & Motivation Related Work System Design CSI Feature Extraction Burst Detection False Alarm Filter Performance Evaluation Conclusion 33
3. False Alarm Filter Target: to reduce the false alarm rate Fact: a single motion instance always lasts a short period Burst Burst true motion! Burst false alarm! 34
Outline Introduction & Motivation Related Work System Design Performance Evaluation Conclusion 35
Experimental Setup Hardware Commercial NICs, APs Software Linux 2.6.38 kernel, Matlab, Python Intel WiFi Link 5300 802.11n Router Testbeds 2 typical indoor environments in HKUST Size 7m 11m (Lab) 32.5m 1.5m (Corridor) Lab Corridor 36
Test outcome: Static Test outcome: Motion Evaluation Metrics Condition: Motion Condition: Static True Positive (TP) Rate False Positive (FP) Rate False Negative (FN) Rate Ignore 37
Effectiveness of FIMD Detection performance w.r.t false alarm (ROC Curve) DR > 70% DR>90% 1% 9% Detection performance w.r.t false alarm is high! 38
Effectiveness of FIMD Influence of the length of sliding window The longer the window length, the less sensitivity 39
Effectiveness of FIMD Influence of the length of sliding window The longer the window length, the lower FP rate 40
Detection Rate CSI vs. RSS Without Narrowband Interference 1 0.8 0.6 0.4 0.2 0 Lab CSI-based RSSI-based Corridor Refer to [PerCom 12] Youssef et. al. RASID system kernel density-based approach 41
False Positive CSI vs. RSS No Motion, with Narrowband Interference Bluetooth 0.7 0.6 CSI-based RSS-based 0.5 0.4 0.3 0.2 0.1 0 SW=5 SW=8 SW=11 SW=14 SW=17 SW=20 42
Outline Introduction & Motivation Related Work System Design Performance Evaluation Conclusion 43
Conclusion We presented a novel device-free indoor motion detection system with commodity hardware in large indoor scenarios. We leverage both the frequency diversity and temporal stability of CSI to enhance detection performance. Experimental results show that CSI-based approach is superior to RSS-based approach in RF domain. 44
Thanks. Questions? jxiao@cse.ust.hk PhD Candidate @ Hong Kong University of Sci.& Tech.