QG6PDUW&LWLHV6\PSRVLXP0DUFK8QLYHUVLW\RI%DKUDLQ

Smart Emergency Response System for Road Accidents with

Automatic Accident Detection

Aditya Shubham*,Ankan Kumar Deogharia*,Ajay Bundela*, Anurag Chaturvedi*,

Lokesh Sharma*, Sumit Srivastava*, Ayush Kshitij+

* Manipal University Jaipur,India

+Delhi Public School Varanasi, India

Keywords: Accident detection, Android, Smart-phone , the same significantly. But to accelerate process of adoption
Emergency of such system, the hardware needs to be portable and mass
-adopted which at present is smartphone. Smartphone with its
Abstract on-board sensors is capable of detecting accelerations that are
higher than that experienced in day-to-day activities even while
Almost four hundred people lose their lives per day in road ac- they might not detect accelerations that are experienced during
cidents in Indian subcontinent alone [3]. We know that the first serious accidents. This paper finds a threshold acceleration that
few minutes play a vital role in deciding the fate of the victim is seldom experienced in day-to-day life while it is definitely
involved in the accident. The Emergency Response Support experienced during the road accidents. We also describe how
System (ERSS) around the world are now improving their in- our threshold is designed to prevent false positives and how
frastructure to find the accurate location of the distress call. On this system can be used to provide analytics data to government
the other hand, sometimes it is not possible for the victim to agencies for making policies and infrastructure keeping safety
call the emergency services. Till now, most of the accident de- in mind.
tection and response mechanisms focused on using specialised
hardware solutions to detect the crash with a few exceptions. This paper contributed the following
This paper presents a way of detecting road accidents through
a smartphone that avoids false positives. An Android • First, solutions to key challenges associated with detect-
smartphone-based prototype architecture of our accident detec- ing traffic accidents, such as preventing false positives.
tion system along with empirical results are being presented.
• Second, a porotype architecture of smartphone-based ac-
cident detection system and empirical analyse and its
1 Introduction ability to resist false positives as well as its capabilities
It has been almost a decade since the beginning of smartphone for accident reconstruction.
era that started by launch of iPhone. It has been one of the • Third, discuss on the process of smartphone-based acci-
most empowering inventions that mankind has ever seen. By dent detection that reduce overall traffic congestion and
2016 around 2.1 billion smartphones were active and during increase the preparedness of emergency responders.
2018 that number has grown to 5.135 billion that helped in
pulling 4.021 billion people online [5][1]. Smartphone is the 2 Related Works
most advanced package of sensors thats available to almost
two- thirds of global population. Nearly 1.25 million people Detecting accidents automatically using a dedicated device has
die in road crashes each year, on average 3,287 deaths a day been interest of academia since a long time. Proposed solu-
in India [3]. While this might not be a big media event, tions for using smartphones for detecting and communicating
but it is more than the number of causalities in air crash or road accidents have existed ever since smartphones came in
mass shootings. A considerable number of people die due to mass market. Iyyappan et al. [14] presented an approach to
slow response to the emergency situation [19]. Automated detect and communicate accidents using a dedicated device fit-
car accident detection can save lives by decreasing the time ted with relevant sensors and microprocessors. Ali et al. [9]
required for information to reach emergency responders. in another paper used a dedicated device for accident detection
This research paper provides making emergency services but relayed the data to smartphone for communication of the
accessible to the victim automatically and making victims in- accident. In 2011,Thompson et al. came up with WreckWatch
formation available to emergency respondent while informing [21], an application that detected accidents using smartphones
family members and the institution from which the person is and provided situational awareness to the emergency respon-
related to about the emergency. dents. They used acceleration filter and microphone filter to
Developing a comprehensive emergency response system for detect the road accidents. However, amid privacy and security
road accidents can help in reducing fatalities caused due to concerns using microphone filter in background is not such a

Authorized licensed use limited to: University of Exeter. Downloaded on July 13,2020 at 12:16:16 UTC from IEEE Xplore. Restrictions apply.
 QG6PDUW&LWLHV6\PSRVLXP0DUFK8QLYHUVLW\RI%DKUDLQ

good idea. Thus, our paper stresses on using only acceleration

filter to detect road accidents. The acceleration filter is high
enough to be difficult to produce manually and low enough to
detect accident. In rare event of false detection of accident, an
alarm starts ringing which gives one minute to user to cross out
the detection of accident and if user does not respond for one
minute it is assumed that accident has happened.

3 Methodology
Android today is used in 84.8 % of smartphones globally [8].
That made android the first choice of deployment for the Emer-
gency Response System for Road Accidents. Almost, all the
android phones have five major sensors Microphone, Camera,
GPS/GLONASS, Proximity Sensor and 3-axis Accelerometer
with sensitivity of 16G [2]. Majorly only two phenomena were
of interest to classify an event into a serious accident. First one
was to look for unusual spikes in acceleration and the second Figure 1. Sign In/Up Screen .
thing to look out was to analyse the sound produced just after
the first event. But amidst privacy concerns [15] the given sys- can help us to detect accidents. But even crucial is to develop
tem is not using the second factor. However, even in absence of a system that does not work with false positives. Phones are
microphone data, accelerometer can predict the accident event subject to fall and heavy braking should not be detected as an
nearly as accurate. accident.

3.1 Android Client

The features discussed above have been integrated in an An-

droid App [17] called Safely (as shown in Figure 1) which is
designed to both satisfy the purpose of detecting accident ac-
curately and feel intuitive to user. It starts with password less
login using Facebook Account Kit [18]. Then there is a small
signup form (Figure 2) which takes critical health and profile
information and sends them to a secure Firebase Server [16]. Figure 6. Flowchart of events once emergency situation is trig-
That opens up a start emergency screen where software button gered
can be pushed to activate emergency situation (as shwon in 4).
Login event also triggers an activity that triggers emergency
3.3 Hardware Used
situation by long pressing power key. User can go to settings
and manually turn on Automatic Accident Detection to activate All experiments were performed on HMD Globals Nokia 7
automatic accident detection module in background. Plus device. It comes with Snapdragon 660 processor coupled
with 4 GB RAM [6]. It comes equipped with Proximity Sen-
3.2 Challenges in Detecting Accidents Automatically sor, Accelerometer, Gyroscope, Microphone and Magnetome-
ter. Its accelerometer can detect accelerations upto 16G.
Modern vehicles come equipped with ECUs that have on-board
sensors and capabilities to detect the accidents [13]. But the 3.4 Solution Approach
most difficult challenge comes in developing a scalable solu-
tion that can interface with a wide variety of ECUs. It also In modern world theres a global trend of increasing smartphone
hardly makes any sense economically to make a dedicated penetration [7], a growing number of people are going to have
hardware for detecting accidents automatically. So, smart- an array of sensors in their pocket. These sensors essentially
phones will hardly be able to detect certain metrics about num- empower us to detect a lot of patterns. It is also possible to
ber of times a vehicle is hit, data about airbag deployment, an- predict the accidents automatically using the same. However,
gle of impact etc. Also, it unrealistic for a user to connect their there are many challenges associated with using smartphone to
smartphones to ECUs before every drive. That makes it almost detect accidents. There are lot of cars that already come within
impossible to get significant data from vehicles onboard Event built accident detection and emergency response systems [12].
Data Recorders [4] and that problem is magnified even further They are specifically designed for this case and predict acci-
due to the lack of common interface for connecting to EDRs dents with high accuracy. However, it is limited to less than ten
and sometimes lack of EDRs in some cars. So, using smart- percent of total vehicles. But smartphones are used in various
phones sensors to detect high accelerations and decelerations other circumstances and are likely to raise false alarms. Thus,

Authorized licensed use limited to: University of Exeter. Downloaded on July 13,2020 at 12:16:16 UTC from IEEE Xplore. Restrictions apply.
 QG6PDUW&LWLHV6\PSRVLXP0DUFK8QLYHUVLW\RI%DKUDLQ

Figure 3. Enter OTP

Figure 2. Enter Mobile Number where,

F = average impact force
m = mass of the object
eliminating false positives due to situations like dropping the v = initial speed of the object
phone is of prime importance. Also, the orientation of phone d = distance travelled during collision
might affect the accident detection. Of course, the interaction
with the Electronic Control Unit frequently is quite tiresome
and not feasible. There is also a need provide the responder
with situational awareness and criticality of the accident to the In case of accident, extending the distance moved during
responders. the collision drastically reduces the average impact force. Al-
We need to develop a comprehensive system despite these chal- ternatively, we can write the above equation as:
lenges and a universal challenge of conserving the battery as
much as possible.
In this case, user has to activate automatic accident detection
module in app before starting the drive. When an accident oc-
curs, phone being in the vehicle experiences same forces as
vehicle itself [10]. Then if at any instant the acceleration goes v∗m
F = (2)
above 13G [20], accident is detected, and a message is sent to t
the emergency contact selected by the user during sign in and a
contact from users workplace along with the users live location.

3.4.1 Impact during the accident

Thus, from above equation we can conclude that extending the
Force of impact is the force generated when objects meet. So, time of collision decreases the average impact force. This can
its the force that vehicles experience during accidents. It can be easily visualized by considering an example where a man
be derived using law of conservation of energy. If a car is jump from a certain height to the ground and to a trampoline.
moving, then it possesses kinetic energy that reduces to zero In case of trampoline the man experiences less average impact
after the accident. Thus, change in kinetic energy must be force due to extended time of collision. This is analogical to
compensated by the work done by the impact force to fulfil the vehicular crashes where extending the time and/or distance re-
law of conservation of energy. duces the impact as a result of crash. In fact, extending the time
Hence, of collision is the main principle of working of seat belts. So,
to understand this impact force let us consider Table 1 given
v2 below to develop an intuition for understanding the following
F =m∗ (1) Impact Force vs Speed graph as shown in Figure 9.
2∗d

Authorized licensed use limited to: University of Exeter. Downloaded on July 13,2020 at 12:16:16 UTC from IEEE Xplore. Restrictions apply.
 QG6PDUW&LWLHV6\PSRVLXP0DUFK8QLYHUVLW\RI%DKUDLQ

S.no. Total Weight of Passengers (Kg) Car Speed (Km/h) Average Impact force (kN) Stopping Time (ms) Deceleration (g)
1 180 30 31.25 48 18
2 180 50 86.8 29 49
3 180 70 170 20.6 96
4 180 90 281 16 159
5 180 110 420 13 238
6 180 130 587 11 332
7 180 150 781 9.6 443

Table 1. Variation in Average impact forces w.r.t. Speed

Figure 5. Nearest Hospitals

Figure 4. Landing page with virtual button

below are the filters used to eliminate false positives:

• GPS consumes a substantial power. Therefore, the au-

tomatic accident detection module is required to be acti-
vated manually to run in background.

• Normal phones can detect acceleration upto 16Gs. Dur-

ing free fall accelerometer cannot detect acceleration of
more than 3Gs and even throwing the phone hard will
make the phone experience a maximum acceleration of
just 9Gs. While some of the airbags deploy at accelera-
tions of upto 60G [11], 13G could be a safe operational
limit to detect accidents. While, airbags experience the
same acceleration/deacceleration as that of vehicle itself,
riders inside the car might experience much lower ac-
celeration/deacceleration due to seat belts, thus, making
mobile phone to experience less acceleration. Hence,
Figure 7. Impact force vs Speed 13G could be a safe operational limit to detect accidents
which is much lower than 60G used in airbags.
3.5 Eliminating False Positives
• Heavy braking from 105 km/hr to 0 km/hr also does not
The major goal of this application was to eliminate false posi- produce acceleration of more than 10G for the person
tives due to heavy breaking or fall and conserve battery. Given having mobile phone in the pocket.

Authorized licensed use limited to: University of Exeter. Downloaded on July 13,2020 at 12:16:16 UTC from IEEE Xplore. Restrictions apply.
 QG6PDUW&LWLHV6\PSRVLXP0DUFK8QLYHUVLW\RI%DKUDLQ

4 Empirical Results 5 Conclusion

Avoiding false positives is a major challenge when detecting In this paper, we propose a novel way of detecting accidents
car accidents with mobile devices. To study the potential for that does not use microphone in background and detects the
false positives, we did two demo experiments designed to sim- accident using accelerometer of the smartphone. The goal of
ulate environment that trigger accelerations whose values could this research was to develop an Emergency Response System
be interpreted as car accidents. For the first test, the mobile for Road Accidents thats portable, accessible and avoids false
device was dropped from ear height in the drivers seat of a positives. This paper also describes how filters are used to ef-
car. Another scenario that could potentially create a false pos- fectively remove false positives and how force generated even
itive is a sudden stop. This test was performed in a car by in very heavy braking is less than the filter. However, this sys-
reaching a speed of approximately 105 kmph and engaging in tem does suffer from limitations especially if the phone is dam-
a sudden stop. The test results are approximate as the exact aged in road accidents. But nevertheless, the benefits of such a
speed was unknown and braking pressure was also not exactly system go beyond just detecting road accidents. The data from
known. Figure 8 below shows the acceleration experienced on accidents such as location of accident, age group of victims,
each axis during the stop. As described above, because the timings of accidents etc can be of extensive use to the govern-
smartphone remained stationary relative to the vehicle, it ex- ments for improving roads and infrastructure that are safer for
perienced the same forces as the vehicle. In this instance, the the public.
acceleration experienced by the smartphone was less than that
experienced during the fall. This result is attributed to the fact
that although the stop was sudden and forceful, the car (and
References
consequently the smartphone) came to a rest over a duration of [1] Global digital 2018 report.
time that was longer than during the drop test. In other words, https://wearesocial.com/us/blog/2018/01/global-digital-
the change in velocity was greater but the actual acceleration report-2018. Accessed on: 2019-02-20.
was less because the change occurred over a longer duration of
time. [2] Report by android.com.
https://source.android.com/devices/sensors/sensor-types.
Accessed on: 2019-02-20.

[3] Report by bbc on road accidents in india,2015.

https://www.bbc.com/news/world-asia-india-36496375.
Published on: 2016-06-10.

[4] Report by national highway traffic safety admin-

istration, united states department of transporta-
tion. https://www.nhtsa.gov/research-data/event-data-
recorder. Accessed on: 2019-02-20.

[5] Report by statista.com on number of smartphone

users worldwide from 2014 to 2020 (in billions).
Figure 8. Time vs Acceleration plot during heavy braking https://www.statista.com/statistics/330695/number-of-
smartphone-users-worldwide/. Accessed on: 2019-02-20.

[6] Report on device specification nokia 7 plus.

https://www.nokia.com/phones/eni nt/nokia − 7 −
plus. Accessedon : 2019 − 02 − 20.

[7] Report on smartphone sales statics.

https://www.statista.com/topics/840/smartphones/.
Accessed on: 2019-02-20.

[8] Smartphone market share- a report by idc.

https://www.idc.com/promo/smartphone-market-
share/os. Accessed on: 2019-02-20.

[9] Asad Ali and Mohamad Eid. An automated system for ac-
cident detection. In 2015 IEEE International Instrumen-
Figure 9. Acceleration plot during sudden fall tation and Measurement Technology Conference (I2MTC)
Proceedings, pages 1608–1612. IEEE, 2015.

Authorized licensed use limited to: University of Exeter. Downloaded on July 13,2020 at 12:16:16 UTC from IEEE Xplore. Restrictions apply.
 QG6PDUW&LWLHV6\PSRVLXP0DUFK8QLYHUVLW\RI%DKUDLQ

[10] Haimd M Ali and Zainab S Alwan. Car accident de- with smartphones. Mobile Networks and Applications,
tection and notification system using smartphone. LAP 16(3):285–303, 2011.
LAMBERT Academic Publishing Saarbrucken, 2017.
[11] Jo Barnes, Andrew Morris, Brian Fildes, and Stuart V
Newstead. Airbag effectiveness in real world crashes.
2001.
[12] Manuel Fogue, Piedad Garrido, Francisco J Martinez,
Juan-Carlos Cano, Carlos T Calafate, and Pietro Man-
zoni. Automatic accident detection: Assistance through
communication technologies and vehicles. IEEE Vehicu-
lar Technology Magazine, 7(3):90–100, 2012.
[13] Varsha Goud. Vehicle accident automatic detection and
remote alarm device. International Journal of Reconfig-
urable and Embedded Systems, 1(2):49, 2012.
[14] Mr S Iyyappan and Mr V Nandagopal. Automatic acci-
dent detection and ambulance rescue with intelligent traf-
fic light system. International journal of advanced re-
search in electrical, electronics and instrumentation en-
gineering, 2(4):1319, 2013.
[15] Asma Khatoon and Peter Corcoran. Privacy concerns on
android devices. In 2017 IEEE International Conference
on Consumer Electronics (ICCE), pages 149–152. IEEE,
2017.
[16] Chunnu Khawas and Pritam Shah. Application of fire-
base in android app development-a study. International
Journal of Computer Applications, 975:8887, 2018.
[17] Jianye Liu and Jiankun Yu. Research on development
of android applications. In 2011 4th International Con-
ference on Intelligent Networks and Intelligent Systems,
pages 69–72. IEEE, 2011.
[18] Michitomo Morii, Hiroki Tanioka, Kenji Ohira, Masahiko
Sano, Yosuke Seki, Kenji Matsuura, and Tetsushi Ueta.
Research on integrated authentication using passwordless
authentication method. In 2017 IEEE 41st Annual Com-
puter Software and Applications Conference (COMP-
SAC), volume 1, pages 682–685. IEEE, 2017.
[19] Stefan Rauscher, Georg Messner, Peter Baur, Jeffrey Au-
genstein, Kennerly Digges, Elana Perdeck, George Ba-
houth, and Oliver Pieske. Enhanced automatic collision
notification system-improved rescue care due to injury
prediction-first field experience. In The 21st Interna-
tional Technical Conference on the Enhanced Safety of
Vehicles Conference (ESV)-International Congress Cen-
ter Stuttgart, Germany, pages 09–0049, 2009.
[20] Dennis F Shanahan. Human tolerance and crash surviv-
ability. Pathological Aspects and Associate Biodynamics
in Aircraft Accident Investigation, 2004.
[21] Jules White, Chris Thompson, Hamilton Turner, Brian
Dougherty, and Douglas C Schmidt. Wreckwatch:
Automatic traffic accident detection and notification

Authorized licensed use limited to: University of Exeter. Downloaded on July 13,2020 at 12:16:16 UTC from IEEE Xplore. Restrictions apply.