Sensing Phone Use of Motorcycle Drivers

Transcription

1 Globecom Wireless Networkig Symposium Sesig Phoe Use of Motorcycle Drivers Jyh-Cheg Che 1, Chu-Feg Wu 2, Wei-Ho Chug 2, Pig-Fa Ho 1 1 Departmet of Computer Sciece, Natioal Chiao Tug Uiversity, Hsichu, Taiwa 2 Research Ceter for Iformatio Techology Iovatio, Academia Siica, Taipei, Taiwa Abstract Due to safety reasos, usig mobile phoes while drivig is prohibited i may coutries. Research has also show that motorcycle riders are 20 times more likely to be killed i a crash tha vehicle occupats. Therefore, it is more critical to restrict the use of mobile phoes of motorcycle drivers tha car drivers. There are some studies that focus o how to distiguish phoe use betwee a driver ad other passegers i a car. The techiques used for cars, however, are ot always applicable to motorcycles. I this paper, we propose a way to detect phoe use of motorcycle drivers. By usig two low-cost Bluetooth emitters, mobile phoes of the driver ad the passeger ca measure the sigal stregths ad decide their locatios. We have coducted extesive experimets with various smartphoes. The results show that o average we ca achieve 96% accuracy. Idex Terms Drivig safety, sesig driver phoe, smartphoe, motorcycle, scooter, Bluetooth, locatio classificatio I. INTRODUCTION Accordig to Goverors Highway Safety Associatio, mobile phoe use ad textig are two of the most commo distractios for drivers which associated with may automobile accidets [1]. A study doe by Virgiia Tech Trasportatio Istitute also shows that sedig or receivig a text takes a driver s eyes away from the road for a average of 4.6 secods. This equates to drivig blid over the legth of a football field at 55 mph [2]. Aother study doe by Natioal Safety Coucil shows that 21 percet of crashes or 1.1 millio crashes i 2010 ivolve talkig o hadheld ad hads-free mobile phoes, ad a additioal 3 percet or more crashes or a miimum of 160,000 of crashes i 2010 ivolve text messagig. Due to safety reasos, it is prohibited to use mobile phoes while drivig i may coutries ow [3] [5]. Although usig a hads-free phoe while drivig is allowed i some coutries, studies have idicated that usig a hads-free phoe while drivig is ot ecessary to be safer tha usig a hadheld phoe [6] [8]. I order to eforce the laws related to mobile phoe use, there are some techical approaches that focus o how to distiguish phoe use betwee a driver ad other passegers for four-wheel motor vehicles 1 so the use of driver s phoe is restricted automatically [9], [10]. Although there are some solutios which may restrict mobile phoe use automatically for drivers of cars, those solutios are ot applicable to two-wheel vehicles icludig mopeds, autobikes, scooters ad motorcycles 2. Research has idicated that motorcycle riders are 20 times more likely to be killed i 1 I this paper, we use cars as a geeric term for all types of four-wheel motor vehicles, such as sedas, vas, trucks, ad buses. 2 Thereafter, we use motorcycles to represet two-wheel motor vehicles icludig mopeds, autobikes, scooters ad motorcycles discussed i this paper. a crash tha car occupats [11]. Accordig to the research ote of the Natioal Highway Traffic Safety Admiistratio, U.S. Departmet of Trasportatio, there were 4,469 motorcyclists killed ad 90,000 ijured i traffic crashes i 2009 [12]. There were 4,502 killed ad 82,000 ijured i 2010 [12]. I Taiwa, approximately 67% of the registered vehicles are scooters ad motorcycles. A average of 1.5 persos ows oe scooter [13], [14]. Statistics also shows that more tha 91% of ijuries ad more tha 79%of deaths i traffic accidets are cotributed by those ivolvig scooters [15]. Experimets idicate that both textig ad phoe use while drivig are as dagerous as druk drivig [16], [17]. It may ot easy for people livig i a coutry without may motorcycles to imagie how to use mobile phoe whe ridig a motorcycle. A commo way is to use oe had to hold the phoe to the ear with the other had o the hadgrip to cotrol the motorcycle. The other way is to cradle a phoe betwee the head ad shoulder. Some people eve isert the phoe ito helmet so the helmet fuctios as a hads-free phoe holder as show i Fig. 1. For textig ad browsig Iteret, a most commo way is to drive with oe had ad use the other had to operate the phoe. Some people eve ca use two hads to text while still drivig. A real case happeed i Austria as show i Fig. 2 which is from the ews i [18]. The rider was caught o video [19] ad lost his licese due to this dagerous behavior. Because it is extremely dagerous, it is illegal to use a had-held or hads-free mobile phoe while ridig a motorcycle i may coutries [5], [20], [21]. There are some solutios which may detect ad restrict mobile phoe use automatically for car drivers. However, motorcycles ad cars are differet i may ways. Specifically, we cosider that there are two seats at most i a motorcycle while there are four seats or more i a car. A commo way to prohibit phoe use is to determie whether the motor vehicle is movig by usig Global Positioig System (GPS) or other meas [22], [23]. Oce the speed is over a certai threshold, a APP i the phoe the block icomig/outgoig calls, texts, ad other messages [24] [28]. Most of the products i the market use this techique. This approach, however, ca ot distiguish driver s phoe ad passeger s phoe. It will block the phoe use of the driver ad all passegers. A most otable solutio is to use the left ad right (ad if possible, frot ad rear) speakers i cars to determie which car seat a phoe is beig used [9]. The solutio is ot applicable to motorcycles because most motorcycles do ot have ay speaker. I additio, its successful rate is oly aroud 90%. There are also other techiques proposed for this problem [10], /14/$ IEEE 4697

2 Globecom Wireless Networkig Symposium (a) (b) A Fig. 2. Usig two hads to text (a real case happeed i Austria) II. D ESIGN C HALLENGES Our mai purpose is to detect the mobile phoe use of the motorcycle driver without affectig the passeger. I this sectio, we delieate the major challeges for our desig. 1) The first challege is that the driver ad passeger are very close to each other i a motorcycle. We measured the seat legth of various types of motorcycles ad scooters. The average legth from two far eds of driver seat ad passeger seat is aroud 63 cm. If we adopt a positio techique [30] [33] to solve the problem, it must be able to achieve a accuracy less tha 63 cm. It is E D F H I Locatios of phoes ad Bluetooth emitters well uderstood that GPS caot achieve such accuracy. Some positioig techiques may meet the requiremet. However, they are either too expesive or ot applicable to motorcycles. 2) A driver ad passeger may put the mobile phoe i differet places, for example i the pocket of shirt or pats. Fig. 3 (c) illustrates that a phoe may be put i positios E, F, G, H, I. The solutio must be able to work correctly o matter where the driver ad passeger put the phoe. 3) A two-wheel motor vehicle usually is much cheaper tha cars. This is especially true for scooters. A ew dailyused scooter i some Asia coutries ca cost as low as $1,000 US dollars. It does ot make sese to adopt the techiques which require expesive equipmet. Therefore, the solutio must be cost-effective. 4) The solutio must be simple ad easy to deploy. It should be uobtrusive ad should ot chage the appearace of the motorcycle. Especially, it should ot icrease the size of the motorcycle. For istace, istallig multiple speakers i the frot ad rear of a motorcycle [9] or a specialized atea [10] is ot preferable. Helmet fuctios as a hads-free phoe holder [29]. However, a driver ad passeger i a motorcycle are very close to each other. It is very challegig to distiguish the driver s phoe ad passeger s phoe is such a small distace. We have doe our best to search related papers ad patets i public domai. To the best of our kowledge, there is o such solutio for motorcycles. We are the first proposed a solutio to detect ad restrict mobile phoe use for motorcycle drivers. A passeger s mobile phoe will ot be affected ad still ca be used. Besides, our proposed solutio is very costeffective. We oly eed to istall two Bluetooth (BT) emitters i the frot ad back of a motorcycle. Our solutio is very simple. The results, o the other had, are surprisigly good. Our experimetal results show that o average we ca achieve 96% accuracy. C B Fig. 3. Fig. 1. (c) G III. S YSTEM D ESIGN I this sectio, we preset our desig. Although there are may challeges, we maage to come out a very simple, costeffective, ad accurate solutio. Our ultimate goal is to distiguish driver s phoe ad passeger s phoe. It is ot to calculate the distace betwee the two phoes. I additio, driver s phoe ad passeger s phoe may be carried differetly as discussed earlier. Therefore, we do ot adopt the positioig ad localizatio techiques [30] [33]. Our solutio is very simple. We oly istall two BT emitters i the frot ad back of the motorcycle as that show i Fig. 3 (a). The BT emitter provides the iformatio icludig BT ID ad Received Sigal Stregth (RSS) which eables the APP i the phoe to decide whether the phoe is i driver s positio. If the sigal stregth received from the frot BT emitter is higher tha the sigal stregth received from the back emitter, it is a driver s phoe. Otherwise, it is a passeger s phoe. To make the sigal stregth i passeger seat lower, our techique is surprisigly simple. Our isight is that we use huma body as a sigal-blockig obstacle. People are actually quite good at blockig BT sigals. Huma body does a great job i absorbig radio waves at the frequecies used by BT because water costitutes aroud 60% of the huma body [34]. 4698

3 Globecom Wireless Networkig Symposium (a) (b) (a) (b) Fig. 4. (a) The RSSs for the phoe at positio E. (b) The RSSs for the phoe at positio F. (c) Therefore, the useful rage of the BT emitter will be reduced. I our desig, we use the driver as the block obstacle. Thus, the problem is solved easily. Ituitively, we ca istall oly oe BT emitter i the frot of the motorcycle to achieve the same purpose. However, the phoe will receive oly oe sigal ad wo t be able to make a proper decisio. Although the RSS measured i driver s phoe geerally will be higher tha that i the passeger s phoe, it is hard to fid a clear cut poit to distiguish driver s phoe ad passeger s phoe. This is because most likely the driver ad passeger will carry differet phoes made by differet maufactures which use differet BT receivers. At the same positio, differet phoes may measure differet RSSs from the same BT emitter. Therefore, o proper RSS threshold eables the APP to correctly make a decisio. We coducted extesive experimets by usig o BT emitter ad cocluded that usig oe BT emitter is ot eough. Due to space limit, we do ot show the results here. IV. ANALYSIS OF BT LOCATION The mai goal of the desig for the two BT locatios is to provide two differet patters for the driver ad the passeger through the RSS ad the BT ID. The RSS itroduced i [35] ca be modeled by the electromagetic propagatio model λ as S RX = E 0 4d where S RX is the sigal stregth at the Receiver RX, E 0 is the sigal stregth at the Trasmitter, the wavelegth λ is equal to 12.5 cm for BT frequecy bad, ad d deotes the distace betwee the trasmitter ad the receiver. Specifically, we deote the RSS for i-th BT Trasmitter at the phoe of Driver as: λ S D,i = E 0, (1) 4d D,i where d D,i is the distace betwee the phoe of the Driver ad the i-th BT Trasmitter. Similarly, the RSS for i-th BT Fig. 5. (a) The RSSs for the phoe at positio G. (b) The RSSs for the phoe at positio H. (c) The RSSs for the phoe at positio I. Trasmitter at the phoe of the j-th Passeger is deoted by: S Pj,i = E 0 λ 4d Pj,i, (2) where d Pj,i is the distace betwee the phoe of the j-th Passeger ad the i-th BT Trasmitter. Based o above, the RSS will decrease whe icreasig the distace betwee the BT Trasmitter ad the Phoe. We exploit the RSS model to aalyze the two BTs i the motorcycle with oe driver ad oe passeger (i.e., i {1, 2}, ad j = 1). Now we cosider the stadard vehicle model, ad the phoe positios for Driver ad Passeger are show i Fig. 3 (c) deoted by E, F, G, H, ad I. Two BTs, deoted by ad, are set at the frot ad back of the motorcycle. The RSSs correspodig to the driver ad the passeger are show i Fig. 4 ad Fig. 5, respectively. The two RSSs correspodig to ad for the phoe at positios E ad F are computed by Eq. (1). The results are show i Fig. 4. As show i the figure, the RSS measured for is always higher tha the RSS measured for. O the other had, Fig. 5 shows the results for positios G, H, ad I by usig Eq. (2). The RSS of is ever greater tha that of. Based o the aalysis, a phoe ca decide whether it is i driver s positio by measurig the RSSs of ad. I additio, all phoes ca make the decisio idepedetly. V. EXPERIMENTAL RESULTS We use two HL-MD08R-C2A BT emitters to coduct our experimets. The HL-MD08R-C2A supports complete 2.4 GHz radio trasceiver ad basebad. Moreover, it supports BT Class 2 with operatio up to 10 meter rage. A BT emitter 4699

4 Globecom Wireless Networkig Symposium TABLE I SMARTPHONES USED IN EXPERIMENTS Adroid Versio Liux Kerel Soy Xperia Lt26i Motorola XT CM SAMSUNG GT-S A Fig. 6. Average RSS measured at Poit A geerally ca be bought at the price less tha $3 US dollars. I the smartphoe, we implemet a APP which first uses GPS to decide whether the phoe (motorcycle) is movig. If it is, the APP will the measure the RSSs to decide the phoe s positio. If it is i driver s positio, commuicatios icludig phoe calls, text messages, ad Iteret access are blocked. Comig phoe calls will be directed ito voice mail automatically. The phoes we used i experimets are listed i Table I. We first show the results that there is o driver ad passeger, as show i Fig. 3 (a). We the preset the results that there is a driver o the motorcycle, as show i Fig. 3 (b). Fially, we demostrate the results that there are a driver ad a passeger o the motorcycle, as show i Fig. 3 (c). A. No Driver ad Passeger: Poit A Whe the smartphoe is located at poit A, the RSS measured from the frot BT emitter should be stroger tha that from the back BT emitter. As show i Fig. 6, however, we ca see that i some cases the received sigal from the back BT emitter is stroger tha that from the frot BT emitter. This is because the distace is too short. The decrease i RSS is ot that sigificat. B. No Driver ad Passeger: Poit B At poit B, there is still o perso o the motorcycle. The results are show i Fig. 7. Similar to the results at poit A, the experimetal results show that ot all of the received sigals from the back BT emitter are stroger tha those from the frot oe. Sometimes it is hard to distiguish where the smartphoe is. The results i Fig. 6 ad Fig. 7 also show that B C Fig. 7. Average RSS measured at Poit B Fig. 8. Average RSS measured at Poit C D Fig. 9. Average RSS measured at Poit D 4700

5 Globecom Wireless Networkig Symposium G H I -10 Fig. 10. Average RSS measured at Poit G Fig. 11. Average RSS measured at Poit H Fig. 12. Average RSS measured at Poit I differet phoes perform differetly at the same positio. This is because each phoe has differet BT hardware ad driver. This also reflects our statemet i Sec. III that usig oly oe BT emitter caot always distiguish the phoe s positio. C. Driver Oly: Poit C Because the back BT sigal is blocked by the driver at poit C, we ca see that all of the results idicate that the RSS from the frot BT emitter is stroger tha that from the back BT emitter as show i Fig. 8. The correct rate to detect a phoe s positio i a loger time scale is listed i Table II. O average, we have 98.75% correct rate for the experimets here. D. Driver Oly: Poit D Fig. 9 illustrates the results for the phoes at poit D. Still, the back BT s sigal is blocked by the driver so the results show that the received sigal from the frot BT emitter is stroger tha that from the back oe. The correct rate of each phoe i a loger time scale is also listed i Table II. I this test, we have 100% correct rate. Comparig the results i Figs. 6 7 ad Figs. 8 9, we ca see that huma body is a good sigal-blockig obstacle. Next, we tur our attetio to the results whe there are a driver ad a passeger. We coducted our experimets with the phoes put at positios E, F, G, H, ad I, as idicated i Fig. 3 (c). Due to space limit, the results for positios E ad F are oly show i Table II. They are similar to those at positios C ad D. E. Driver ad Passeger: Poit G Fig. 10 ad Table II show the results whe the smartphoe is located at poit G. Because the phoe is i-betwee the driver ad the passeger, both the frot ad back sigals are blocked. The results idicate that the SONY Xperia Lt26i ad oly have 60% ad 55% correct rate, respectively. However, the other two phoes have ear 100% correct rate. We thik the reaso is that the atea desig of differet BT modules is differet. We pla to test more phoes ad provide a list of the results so users ca choose a phoe with good performace. If there is a law to prohibit phoe use whe drivig, the govermet ca list the requiremets of the BT module i a phoe ad eforce the maufactures to comply with the requiremets. F. Driver ad Passeger: Poit H Whe the phoe is put at poit H, the sigal of the frot BT is blocked by the driver ad the sigal of the back BT is blocked partially by the passeger. Fig. 11 shows the experimetal results. Table II idicates that the correct rate is 96%. G. Driver ad Passeger: Poit I At poit I, the frot BT sigal is blocked by both the driver ad the passeger. Moreover, the phoe is very closed to the back BT emitter. Fig. 12 shows that i geeral it is easy to decide that it is passeger s phoe. Table II idicates that the correct rate is 99%. 4701

6 Globecom Wireless Networkig Symposium TABLE II CORRECT RATE C D E F G H I Soy Xperia Lt26i 95% 100% 100% 95% 60% 100% 100% Motorola XT % 100% 100% 100% 95% 90% 95% 100% 100% 100% 100% 55% 100% 100% SAMSUNG GT-S % 100% 100% 100% 100% 95% 100% Average 99% 100% 100% 99% 78% 96% 99% I summary, our proposed solutio has % correct rate whe there is oly a driver o the motorcycle. Whe there are both a driver ad a passeger o the motorcycle, our method have 94.25% correct rate. I all of our experimets, the average of correct rate is 95.71%. VI. SUMMARY Although there are some techiques proposed to distiguish driver s phoe ad passeger s phoe i cars, they are either too costly or ot applicable to motorcycles. For example, a recet study preseted i [9] eeds to use the left ad right (ad if possible, frot ad rear) speakers. Still, its correct rate is oly aroud 90%. I this paper, we propose a very simply but effective way for motorcycles. We oly eed two BT emitters which are very cheap. O the other had, we ca achieve 95.71% correct rate. VII. ACKNOWLEDGMENTS This work was supported i part by the Miistry of Sciece ad Techology of Taiwa uder grat umbers NSC E MY3, NSC E MY2, MOST E MY3, ad MOST E REFERENCES [1] G. H. S. Associatio, Distracted Drivig: What Research Shows ad What States Ca Do, 2011, sfdist.html. [2] F. M. C. S. A. U. D. of Trasportatio, Driver Distractio I Commercial Vehicle Operatios, [3] G. H. S. Associatio, Cell Phoe ad Textig Laws, Mar. 2013, http: // laws.html. [4] Usig your mobile phoe while drivig, 2013, gov.uk/usig-your-mobile-phoe-while-drivig. [5] Usig mobile phoes whe drivig: the law, Feb. 2013, gov.uk/usig-mobile-phoes-whe-drivig-the-law. [6] Y. Ishigami ad R. M. Klei, Is a hads-free phoe safer tha a hadheld phoe? Safety Research, vol. 140, pp , [7] J. K. Caird, C. R. Willess, P. Steel, ad C. Scialfa, A Meta-Aalysis ofthe Effects of Cell Phoes o Driver Performace, Accid Aal Prev., vol. 40, pp , [8] P. J. Treffer ad R. Barrett, Hads-free mobile phoe speech while drivig degrades coordiatio ad cotrol, Trasportatio Research: Part F: Traffic Psychology ad Behaviour, vol. 7, pp , [9] J. Yag, S. Sidhom, G. Chadrasekara, T. Vu, H. Liu, N. Ceca, Y. Che, M. Gruteser, ad R. P. Marti, Sesig Driver Phoe Use with Acoustic Ragig through Car Speakers, IEEE Tras. Mobile Comput., vol. 11, pp , Sep [10] J. M. Rodriguez-Ascariz, L. Boquete, J. Catos, ad S. Ortega, Automatic system for detectig driver use of mobile phoes, Trasportatio Research Part C: Emergig Techologies, vol. 19, pp , Aug [11] Roads ad M. Services, NSW Cetre for Road Safety: Motorcycles, 2012, idex.html. [12] N. H. T. S. Admiistratio, Traffic Safety Facts: Research Note, Feb. 2012, [13] M. of trasportatio ad commuicatio R.O.C, Motor Vehicle Registratio, Ja. 2012, &fuid=e3301. [14] D. of Household Registratio M.O.I., Household statistics data aalysis, Feb. 2012, [15] I. of Trasportatio. MOTC, Relatioship betwee accidets ad cause of death, 2011, [16] Textig While Drivig: How Dagerous is it? com/features/textig-while-drivig-how-dagerous-is-it. [17] Cell Phoe vs. Druk Drivig MiiMyth, [18] Look o hads: Textig at 80km/h o a motorbike, May 2012, http: //tiyurl.com/bldl45s. [19] Textig motorbike rider loses licese, May 2012, a7gso8a. [20] Roads ad M. Services, Learer rider licece, 2012, 3w48y. [21] Regulatory situatio i the Member States regardig Fial Report, 2009, Nomadic\ Devices.pdf. [22] G. Chadrasekara, T. V. A. Varshavsky, M. Gruteser, R. P. Marti, J. Yag, ad Y. Che, Trackig vehicular speed variatios by warpig mobile phoe sigal stregths, i Proc. It l Cof. Pervasive Computig ad Comm. (PerCom 11), North Bruswick, NJ, USA, Mar. 2011, pp [23] D. Gudlegard ad J. Karlsso, Hadover Locatio Accuracy for Travel Time Estimatio i Gsm ad Umts, Proc. IEEE Itelliget Trasportatio Systems Cof. (ITSC 09), vol. 3, pp , Mar [24] Drivesmart Plus, 2014, [25] Guardia Agel Vehicle Platform, 2014, [26] Key2SafeDrivig App, 2014, [27] Textecutio, 2014, [28] txtblocker, 2014, [29] J. She, I-Vehicle Driver Detectio Usig Mobile Phoe Sesors, 2011, Chu.pdf. [30] L. Girod, M. Lukac, V. Trifa, ad D. Estri, The desig ad implemetatio of a self-calibratig distributed acoustic sesig platform, i Proc. Fourth It l Cof. Embedded Networked Sesor Systems (Sesys 06), New York, NY, USA, 2006, pp [31] A. Harter, A. Hopper, P. Steggles, A. Ward, ad P. Webster, The aatomy of a cotext-aware applicatio, i Proc. of ACM/IEEE Iteratioal Coferece o Mobile Computig ad Networkig (MobiCom), New York, NY, USA, 1999, pp [32] C. Peg, G. She, Y. Zhag, Y. Li, ad K. Ta, BeepBeep: A High Accuracy Acoustic Ragig System Usig COTS Mobile Devices, i Proc. Fifth It l Cof. Embedded Networked Sesor Systems (Sesys 07), 2007, pp [33] N. B. Priyatha, A. Chakraborty, ad H. Balakrisha, The Cricket Locatio-Support System, i Proc. of ACM/IEEE Iteratioal Coferece o Mobile Computig ad Networkig (MobiCom), New York, NY, USA, 2000, pp [34] A. S. Huag ad L. Rudolph, Bluetooth Essetials for Programmers. Cambridge Uiversity Press, Sep [35] C. A. Balais, Atea theory: aalysis ad desig. Joh Wiley & Sos,