1. D ETECTING D RIVER P HONE U SE L EVERAGING C AR S PEAKERS Jie Yang, Simon Sidhom, Gayathri Chandrasekaran, Tam Vu, Hongbo Liu, Nicolae Cecan, Yingying Chen, Marco Gruteser, Richard P. Martin MobiCom 2011 - Sowhat 2011.10.24

2. O UTLINE Introduction System design Experiment Results Conclusion Discussion

3. O UTLINE Introduction System design Experiment Results Conclusion Discussion

4. M OTIVATION I NTRODUCTION M OTIVATION Cell Phone Distraction Hands-free Phone Operation Driving Mode Route incoming calls to voicemail Delay incoming text notification Driving Mode Route incoming calls to voicemail Delay incoming text notification Driver? Passenger?

5. O BJECTIVE I NTRODUCTION O BJECTIVE Classify phone use @ which seat Bluetooth + Audio infrastructure Phone Transmit Signal Car Head Unit Speaker Play Signal Phone Record Processing

6. C HALLENGES I NTRODUCTION C HALLENGES Unobtrusiveness Robustness to Noise and Multipath Computational Feasibility

7. O UTLINE Introduction System design Experiment Results Conclusion Discussion

8. O VERVIEW S YSTEM D ESIGN O VERVIEW Acoustic Relative-Ranging System Avoid Interfere Distinguish speakers

9. R ELATIVE R ANGING S YSTEM D ESIGN R ELATIVE R ANGING

10. B EEP S IGNAL S YSTEM D ESIGN B EEP S IGNAL Unobtrusiveness Robustness to Noise Frequency Android developer phone 2 – 16~18kHz iPhone 3G – 18~20kHz Length – 10ms

11. P ROCESSING S YSTEM D ESIGN P ROCESSING

12. D ETECT B EEP S YSTEM D ESIGN D ETECT B EEP Sequential change-point detection : sequence of recorded audio signal Requirement: known distribution Cumulative sum μ : mean value of noise from t 0 to t 1, time of transmitting > Threshold for W samples

13. P ROCESSING S YSTEM D ESIGN P ROCESSING

14. C LASSIFICATION S YSTEM D ESIGN C LASSIFICATION 2-channel > -5 4-channel if  Front Check Else  Back Check

15. O UTLINE Introduction System design Experiment Results Conclusion Discussion

16. P HONES & C ARS E XPERIMENT P HONES & C ARS Phones Android Developer Phone 2, Phone I iPhone 3G, Phone II Cars Honda Civic Si Coupe, Car I Acura Sedan, Car II 2 channel sound system with fader system

17. P ROCEDURE E XPERIMENT P ROCEDURE Create Beep 4 beep audio files, For { Phone I, Phone II } x { 2 channel, 4 channel } Each file with 4 beeps for each channel Create uniformly distributed white noise  Bandpass filter {16~18, 18~20} kHz Transmit Laptop to car’s head unit via Bluetooth Record Audio recorder app. on phone for offline analysis

18. S CENARIOS E XPERIMENT S CENARIOS Phone I, Car I

19. S CENARIOS E XPERIMENT S CENARIOS Phone II, Car II

20. S CENARIOS E XPERIMENT S CENARIOS Phone I, Car I @ 60 MPH Highway

21. O UTLINE Introduction System design Experiment Results Conclusion Discussion

22. M ETRICS R ESULTS M ETRICS Accuracy Detection Rate False Positive Rate Measurement Error

23. D RIVER VS. P ASSENGER R ESULTS D RIVER VS. P ASSENGER 2-Channel is OK

24. P OSITION A CCURACY R ESULTS P OSITION A CCURACY

25. L EFT VS. R IGHT R ESULTS L EFT VS. R IGHT

26. F RONT VS. B ACK R ESULTS F RONT VS. B ACK

27. R ELATIVE R ANGING R ESULTS R ELATIVE R ANGING

28. C OMPUTATIONAL C OMPLEXITY R ESULTS C OMPUTATIONAL C OMPLEXITY Filtering, Short-time Fourier transform n recorded samples, moving window size After filtering, Average processing time 0.5s for 2-channel 1s for 4-channel

29. O UTLINE Introduction System design Experiment Results Conclusion Discussion

30. C ONCLUSION Driver mobile phone use detection system Leverage existing infrastructure of car speaker for ranging via Bluetooth Acoustic relative-ranging Over 90% detection rate

31. O UTLINE Introduction System design Experiment Results Conclusion Discussion

32. D ISCUSSION Strengthens Clear motivation, system design description Real implement Weakness Different setting for different hardware  not comparable How to set calibration value

33. T HANKS FOR L ISTENING ~