Smartphone-Based Intelligent Driver Assistant: Context Model and Dangerous State Recognition Scheme

  • Igor LashkovEmail author
  • Alexey Kashevnik
Conference paper
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 1038)


The paper proposes the context model and the dangerous state recognition scheme for intelligent driver assistant system. The system is aimed at utilization of smartphone’s front-facing camera and other sensors for dangerous states reignition to prevent emergency and reduce the accidents probability. The proposed context model is divided into following types of contexts, driver context, vehicle context, road context, and environment context. The model shows how the smartphone front-facing camera and sensors as soon as accessible Internet services are used to support the proposed context types. Then, the context-based dangerous state recognition model is presented. The model calculates the computational power of the smartphone and based on this information implements the frame skipping algorithm to reduce the computation complexity for the smartphone processor. The implementation of the proposed frame skipping model shows that for the modern smartphones the complexity is decreased by three times in compare with usual scheme. The proposed context model and dangerous state recognition scheme has been implemented in Drive Safely system that is available in Google Play.


Context-aware driver assistant Dangerous state Context Smartphone Vehicle Driver 

1 Introduction

Road accidents remain the most disastrous of almost any country in the whole world. Drowsiness, distraction or alcohol intoxication of the driver are the most common causes of vehicle related dangerous situations. It should be highlighted that one of the increasingly popular approaches presented in previous scientific researches relies in the development of advanced of driver assistance systems. These safety systems are aimed at reducing road accidents and providing better interaction and engagement with a driver. Some common examples of driver safety technologies for this type of systems are vehicle collision avoidance system, lane keep assistant, driver drowsiness and distraction monitoring and alerting. General use of such systems comprises a certain sequence of commands and can be described in this way: monitoring driver behavior, state of the vehicle or road situation by using different built-in auxiliary devices, including short and long-range radars, lasers, lidars, video cameras to perceive the surroundings; continuous analysis of reading from sensors and determining of dangerous situations while driving; notifying driver about recognized unsafe in-cabin and road situations; and taking control of the vehicle if driver reaction is not sufficient or missing. At the moment, driver safety systems heavily rely on data collected from different in-vehicle sensors. This data can aid to describe changes in the surrounding environment by using both single and multiple sensors together in different situations and, therefore, provide relevant context. According to the definition of Dey [1] the context represents any information that can be used to characterize the situation of a person, place or object, that is considered relevant to the interaction between the user and the application. Some examples of information that can be associated with a driver and environment and describe driver’s situation for a safety driver assistance application are weather conditions, traffic information, vehicle characteristics or physiological parameters of the driver. This kind of information can affect the driving performance in different real-time situations.

There is a certain research and technological gap in implementing the context-aware approaches for better understanding vehicle driver and current environment situation. That is, situation relevant information is not a contributing factor for determining unsafe driving behavior, alerting driver about dangerous road situation, generating context-based recommendations, preventing or mitigating traffic accidents or adapting system for driver needs, taking into its account preferences and competencies. This paper considers the advantages of context utilization for driver assistance application, presents a context model and the idea of implementing context in the application. The design and prototyping of this application have been described in our previous works [2, 3, 4], including the mobile application for Android smartphones1. This paper extends our prior work.

The rest of the paper is organized as follows. Section 2 presents a comprehensive related work in the area of context-aware solutions for safe driving and compares the available solutions. Section 3 describes in detail our context-aware model. The context-aware dangerous state detection algorithm based on camera frame skipping is presented in Sect. 4. The implementation and evaluation are presented in Sect. 5. It is followed by discussion (Sect. 6). Finally, the main results and potential future work are summarized in Conclusion (Sect. 7).

2 Related Work

Firstly, we can distinguish following types of context related to driver assistance systems: driver, vehicle and environment. On one hand, this type of information is provided, in general, by electronic control unit that is already embedded in automotive electronics and controls vehicle’s systems or subsystems. On the other hand, in recent years modern smartphones became a perspective powerful device not only for calls and SMS messages, but also for a variety of diverse tasks including informational, multimedia, productivity, safety, lifestyle, accessibility related applications and many others. Most modern smartphone already have a set of built-in sensors [5], measuring some kind of physical quantity and, in general, include accelerometer, gyroscope, magnetometer, Global Positioning System, light and proximity sensors. Because of its affordable low price, a set of embedded sensors and small sizes, smartphones are gaining popularity for building driver assistance systems at a large scale.

The study [6] estimates the influence of the smartphone based advanced driver assistance system (ADAS) on driving behavior when driver is distracted by mobile social networking applications. The proposed approach utilizes the specially designed driving simulator that replicates the car structure and its parts and includes following third-party components. In order to analyze driver behavior, the eye-tracking system Tobix X120 was used to show where the people are looking exactly. Smartphone was used as a platform for ADAS system and to provide continuously taken images from the front-facing camera to detect the presence of head and eyes of the driver in the scene and alert its through a single beep of 1250 ms. This study showed that using front-facing camera of the smartphone for ADAS aids to track driver behavior, mitigate the distracted tasks and improve the overall driving performance. The data for their research was the driving behavior perceived by the video camera and eye tracking system.

Another research [7] tackles the problem of driver behavior classification for driver assistance systems. In this paper the neuro-fuzzy system is proposed to estimate the driving behaviors based on their similarities to fuzzy patterns. Sensor fusion is used by the neural network for recognizing types of driver maneuvers, including driver’s lane change, left and right turns and U-turn. The initial data source for these maneuvers are raw data readings from accelerometer, gyroscope and magnetometer, that measure changes in velocity, magnetic field and rotation velocity respectively. The output results of the proposed method are two different scores, safe driving score and aggressive driving score. The results of this paper state that the evaluation of driving behavior plays a crucial role in improving driver safety.

Based on the fact that ignoring road signs by a vehicle driver can lead to major accident, this approach [8] is focused on traffic sign recognition using contour analysis approach. The proposed system makes an audible alert for a driver of road signs and, therefore, helping its to avoid a traffic accident. This paper only considers the information of traffic sign boards mounted along the roads for safe driving, that in its turn describes the context relevant for the environment.

Additionally, the vehicle context can provide information about vehicle characteristics and its capabilities in a certain situation scenario. In the paper [9], the developed components utilize on-board diagnostics communication module to acquire vehicle information such as mileage information, sensor data, including fuel usage, velocity, etc. Afterwards it provides information tips about eco-driving and safe-driving. This kind of information can aid to influence the driving style and increase the overall driving safety.

Patent [10] shows a possible integration of data from smartphone sensor devices and vehicle on-board diagnostic system to address a problem of generated pollution. Smartphone sensors used in the study are accelerometer and location-based GPS. On-board diagnostics port provides data about accelerometer, tire pressure, velocity, anti-lock brake condition, temperature and others. Such system can be used in ranking driving style and estimating its fuel economy. Collected sensor data can be applied for further analysis in remote services.

Another study [11] presents an approach that operates driver personal information (age, gender), traffic violations and traffic accident records dataset for furthermore traffic safety assessment. Traffic violations were taken from Public security recommendation standard GA/T 16.30-2017 and some examples of them are parking in the prohibited area, running a red light, speeding, overload, not wearing a seat belt. The performance and accuracy of the proposed data mining framework was evaluated by Kolmogorov-Smirnov chart and the receiver operating characteristic curve. This study is based only on using driver personal information to safety assessment.

One more type of information involved in traffic safety is provided by Green Light Optimal Speed Advisory (GLOSA) systems, allowing to achieve reduced emissions of vehicles and increase the efficiency of traffic flow in the areas of the signalized intersections. One of the studies [12] addresses the performance of this type of systems in terms of fuel consumption, carbon dioxide emissions of vehicles and travel time. Conducted experiments show that in certain situations whereas 400–500 vehicles are going per hour, the developed GLOSA system is able to reduce congestion, enhance the environment state and, therefore, improve the overall driving performance. This system is relied on using data taken from wireless vehicle-to-infrastructure [13] communication systems. The parameters measured in this study are travel time, consumption rate, CO2 emissions for every 60 s, including information about acceleration, velocity and position of the vehicles calculated at step of 0.1 s.

Nowadays, connected-vehicles technology [14] is gaining popularity among researchers and focused on increasing road safety and driving performance. It utilizes dedicated short-range protocol either on-board units already integrated in some vehicles to transmit information messages between vehicles or infrastructure at a predefined rate. Generally, these transmitted safety messages can include information about vehicle unique identifier, its GPS position, speed, time and driving direction. One of the scenarios of using connected-vehicle technology is vehicle platooning [15], that is built with Vehicle-to-Vehicle communication and aids a group of vehicles to travel very closely together. To maintain a vehicle set, each vehicle in a chain requires environmental information and needs share sensing data, including route information, vehicle speed, acceleration, heading direction, future intent actions (e.g. turning, braking, changing lane, etc.), vehicle parameters or driver behavior information like reaction time, so that other vehicles will be notified before attempting maneuver. This technique allows to increase the capacity of roads, road safety and provides steady state of traffic and, therefore, can reduce fuel consumption.

Big scale vehicle systems in terms of growing data flows, comprising sensing data, collecting and sharing information about vehicles, pedestrians, road and weather conditions are built for autonomous vehicles [16]. They utilize a numerous number of modern sensors, including cameras, LIDARs, radars and lasers to facilitate different safety technologies.

Today, user assistants are gaining popularity in many aspects of life. Currently available digital assistants are focused on making user life easier and more comfortable by automating different types of daily tasks. Context-relevant information can be perfectly fitted in vehicle safety systems in a form of intelligent driver assistants [17, 18]. This way driver assistant can provide actionable information for a driver in a certain scenario to increase its driving performance and road safety. Typical use cases for a vehicle driver are context-relevant driving safety tips or recommendations for navigation, eco driving, safe driving using visual format or natural language dialog.

One of the positive aspects of the listed research studies is that they consider different types of context for their needs. But the work of these systems generally relies on the utilization of certain types of information. The use of information relevant for various real scenarios can influence the performance of some work or process, optimize it and potentially propose the new ways of solving tasks.

3 Context Model

Driver behavior is a result of complex interactions between the driver, the vehicle and the environment. According to the above listed studies and the survey of research studies related to the smartphone-based context [19], sensor’s information taken from the smartphone can be categorized into four following groups. Based upon these formed groups the context model for the intelligent driver assistant on the road has been developed as shown in Fig. 1.
Fig. 1.

Context model for intelligent driver assistant on the road.

Driver context includes type of smartphone used by a driver; physiological state like drowsiness, distraction or drunk state; reckless (abnormal) driving actions (parameters) the driver makes on public road; the driving style (eco, normal, aggressive driving); smoothness of driving, in-cabin situation (speaking with passengers, noise level), navigation route provided by some navigation system, personal schedule or calendar of tasks provided by some third party service; reminder list; driver preferences (e.g. music genre, audio level); phone contact list; trip information, including trip elapsed time and distance, start point of trip, destination point of trip; internet connection availability; health related information (e.g. from wearable device) that are pulse or pressure; and driver reaction time.

Vehicle context: the vehicle location (longitude, latitude, altitude), speed (speedometer), fuel level, fuel usage, accelerator pedal position, airbag state, yaw angle, tilt angle, RPM vehicle infotainment system state, tire pressure (flat tire), lights state (switch on/off), vehicle engine status.

Road context: nature of surface, width, conditions, obstacles, road accidents, accident rate, traffic congestion (e.g. traffic volume), road works, road signs (types: guidance, warning, regulation), (e.g. speed limit, driving directions), other road users (e.g. vehicles, pedestrians, cyclists), traffic signals (e.g. red traffic signal), lightning for section of road.

Environment context: weather conditions (e.g. humidity, temperature, pressure, wind speed/direction), weather forecast, current time, nearest POIs (Point of Interest) (location, working hours, food and beverages availability, prices, place to sleep, gas availability).

These formed groups of context allow to describe driver for every real situation.

4 Context-Aware Dangerous State Detection Based on Camera Frame Skipping

The process of continuously recognizing dangerous driving states for each frame taken from the front-facing camera can significantly reduce the battery charge over time and make a noticeable load not only for the whole application, but also affect the operating system of the smartphone. Therefore, the context can be used to address this problem.

Numerous technical improvements had been recently contributed to driver assistance systems to increase road safety and driving performance. This kind of driving safety systems is limited in using personalization and adaptation for a driver. Developed context model can aid system to make context-based decisions while driving by managing driver personal information and situation-relevant information.

There is a significant number of use cases whereas context can improve the overall system performance. A number of research studies proves that modern smartphones [20] can be considered as a standalone platform with integrated cameras, GPS, accelerometer, magnetometer, gyroscope sensors inside and can be efficiently utilized by the drivers to reduce or mitigate the occurring road accidents, improve their safety, driving skills and performance. The essential part of the modern smartphone is a front-facing camera that outputs a set of consecutive image frames. This type of camera is pretty generally applicable in continuously monitoring driver behavior and recognizing dangerous situation detection tasks. Assuming the detection of dangerous situations should work without interruptions for each incoming frame, this process can quickly drain the smartphone battery till the full discharge. To improve the smartphone’s battery usage and reduce the application overall load on the CPU and other OS systems, we propose to consider the following developed context-based algorithm for skipping irrelevant camera frames. The generalized process of processing camera frames is shown in Fig. 2.
Fig. 2.

Camera frame processing on the timeline.

Current research studies and projects show that the time to collision parameter is actively used in building-up monitoring driving behavior systems. This parameter depends on the time, that is used for dangerous state recognition and driver’s reaction time. The driving safety system should place its work in the interval \( t_{s} \) used for detecting dangerous situation \( t_{e\left( i \right)} \) in each frame taken from the front-facing camera. Since modern smartphones do not lack of high-performance computational power, they are able to process a pretty high number of camera frames in a certain period of time, reaching a peak of more than 10 frames per one second. This number of frames is excessively much for predicting and recognizing hazardous driving behavior in a real situation through the interval of 1.5 s.

A number of frames exceeding the high limit of maximum processed frames \( n_{limit} \) can be evenly skipped while recognizing dangerous situation without strong influence on the accuracy of the ongoing process. This is achieved by diving the time of dangerous state recognition in time periods for recognizing unsafe driving behavior and skip time intervals. Skip time interval is a period of time during which driver behavior recognition task can be skipped for the frame received from the camera. This parameter is estimated as the ratio of time needed for dangerous state recognition \( t_{s} \) and max count of processed frames \( n_{limit} \) for the dangerous state, excluding the average processing time of a single frame \( t_{e}^{avg} \).

The full scheme of context-based algorithm for frame skipping while recognizing dangerous state is presented in Fig. 3 and can be described as follows. The goal of the whole algorithm is to estimate the skip time interval that can be further used for choosing only a limited number of frames \( n_{limit} \) that can be submitted for dangerous state recognition task.
Fig. 3.

UML diagram of camera frame skipping while detection of dangerous states.

The preliminary operation of the algorithm is a warming-up action, that skips first “cold” camera frames COLD_F for stabilizing further frame processing calculations. The first task of the algorithm is to estimate the average time to process single frame \( t_{e}^{avg} \) and average skip time interval \( t_{sk} \) this way (1):
$$ t_{sk} = \frac{{t_{s} }}{{n_{limit} }} - t_{e}^{avg} $$
where \( t_{e}^{avg} \) is as an average processing time of single frame. Total time of skipping frames is calculated as the sum of all skip time intervals \( t_{sk}^{sum} \) related to dangerous state (2):
$$ t_{sk}^{sum} = \sum t_{sk} $$

Otherwise, if the average skip time interval \( t_{sk} \) has been already calculated, we check whether the sum of skip time intervals \( t_{sk}^{sum} \) is equals or greater than a fixed skip time interval \( t_{sk} \) and, if it does, we subtract \( t_{sk} \) from \( t_{sk}^{sum} \) and, finally, we proceed to frame recognition. In case, if there is no enough time skip interval for current dangerous state recognition, we skip this camera frame and wait for a new one. This algorithm involves the processing of every frame read from the smartphone camera one after another.

The proposed context-based algorithm can aid driver assistance system to reduce the battery drain and increase the system performance for other vehicle related tasks.

5 Evaluation

The evaluation has been carried out with 15 volunteer drivers of different ages in vehicles and implemented in according to the following methodology. The implementation of the proposed context-based algorithm for frame skipping while recognizing dangerous state has been developed in Kotlin language2 (Listing 1) and, afterwards, tested in the mobile application installed on smartphones. This approach estimates the time recognition, smartphone’s battery drain and performance while determining dangerous driving behavior.
The smartphone camera video stream is created and started to continuously receive preview frames from the front facing camera. The application utilises the developed context-aware algorithm for skipping irrelevant frames to optimize the smartphone’s battery usage and improve the performance of the overall application for other heavy tasks. The applicability and validity was shown in the experiment of the comparison of the time recognition to detect dangerous state in driving behavior with smartphones of different models and manufactures (Fig. 4), where E is a smartphone computing power, n is the number of camera frames being processed, t is the time that the smartphone uses to determine dangerous state in driving behavior. Evaluation of the time required to recognize dangerous state of the driver was conducted provided that the driver’s reaction time is defined as 0.5 s, being a quite common for drivers. The most of the presented smartphones that are mid-range phone or flagships may recognize more than 12 dangerous situations in two seconds that is a high bound of the acceptable range. It means that, for example, Xiaomi Mi 5 smartphone, can skip seven frames and provide nearly the same performance of dangerous state recognition for a driver by processing only five camera frames.
Fig. 4.

Evaluation of time recognition and smartphone’s performance while determining dangerous driving behavior

The proposed algorithm was integrated in the mobile Drive Safely application that is intended for Android-based smartphones. The screenshot of the application (Fig. 5) shows that the application recognized drowsiness dangerous state of the driver and made a warning signal for him in a form of textual messages and audible signals.
Fig. 5.

Screenshot of the Drive Safely application in the moment of recognizing drowsiness dangerous driver behavior utilizing the front-facing camera of the smartphone and alerting driver to pay its attention.

6 Discussion

In this study, we proposed the context-aware approach that accumulates the information about vehicle driver and the current environment situation (context). The context is applicable for monitoring driving behavior, adapting safety system for a driver, and recognizing dangerous situations, early alerting driver about road hazards and preventing or mitigating traffic accidents. This approach eliminates the drawbacks of current research studies and solution by leveraging different types of context. The proposed context-based approach was implemented and evaluated with the aid of Drive Safely mobile application for smartphone. Nevertheless, the presented context-based approach can be also successfully applied in fastest-growing category of advance driver assistance systems, in-cabin cameras focused on monitoring driver facial features and wearable electronic devices, equipped with built-in sensors, tracking driver’s biological measurements while driving. Although advanced driver assistance systems equipped with many high-precision sensors show high accuracy and performance in recognizing driving dangerous situations in different weather conditions, the mobile applications built for smartphones are at much lower price, and are much popular among people in almost every country that is easy to use in every vehicle.

The beneficial effect of the proposed context-based approach is to consolidate the information about driver profile, its preferences and environment situation that would allow driver safety system to increase its performance and accuracy and adapt for driver needs in better personalized way.

7 Conclusion

The paper presents the context model for the intelligent driver assistant on the road. This model was divided into four following groups: driver context, vehicle context, road context and environment context. The developed context model can aid system to make context-based decisions while driving by managing driver personal information and situation-relevant information. To improve the smartphone’s battery usage and reduce the application overall load on the different subsystem of the operation system, we proposed the context-based algorithm for skipping irrelevant camera frames without affecting the accuracy of dangerous driving state recognition. The evaluation of the developed context-aware algorithm of skipping camera frames shows that it allowed to reduce the computation complexity to the smartphone processor by three times in compare with general scheme.

In future work, we expect to collect the dataset of driving statistics in real scenarios involving people of different age with different vehicles to test the proposed context model. We plan to estimate the influence and performance of concrete context parameters on the overall driving safety by utilizing machine learning techniques. In these experiments we intend to use the developed Android-based mobile application Drive Safely aimed at recognizing dangerous behavior and alerting driver to prevent the road accident. As an extension to this work, we consider adding more driving dangerous states, that is drunk driving, aggressive driving, and involving of utilizing non-smartphone external sensor devices, that in its turn can allow to expand the use of proposed context-based approach.




The research is funded by the Russian Science Foundation (project # 18-71-10065).


  1. 1.
    Dey, A.K.: Understanding and using context. Pers. Ubiquitous Comput. 5(1), 4–7 (2001)CrossRefGoogle Scholar
  2. 2.
    Smirnov, A., Kashevnik, A., Lashkov, I., Hashimoto, N., Boyali, A.: Smartphone-based two-wheeled self-balancing vehicles rider assistant. In: Proceedings of the 17th IEEE Conference of the Open Innovations Association FRUCT, Yaroslavl, Russia, pp. 201–209 (2015)Google Scholar
  3. 3.
    Smirnov, A., Kashevnik, A., Lashkov, I., Baraniuc, O., Parfenov, V.: Smartphone-based dangerous situation identification while driving: algorithms and implementation. In: Proceedings of the 18th IEEE Conference of the Open Innovations Association FRUCT, Finland, pp. 306–313 (2016)Google Scholar
  4. 4.
    Lashkov, I., Smirnov, A., Kashevnik, A., Parfenov, V.: Ontology-based approach and implementation of ADAS system for mobile device use while driving. In: Proceedings of the 6th International Conference on Knowledge Engineering and Semantic Web, Moscow, CCIS 518, pp. 117–131 (2015)Google Scholar
  5. 5.
    Fazeen, M., Gozick, B., Dantu, R., Bhukhiya, M., Gonzalez, M.C.: Safe driving using mobile phones. IEEE Trans. Intell. Transp. Syst. 13(3), 1462–1468 (2012)CrossRefGoogle Scholar
  6. 6.
    Dumitru, A.I., Girbacia, T., Boboc, R.G., Postelnicu, C.-C., Mogan, G.-L.: Effects of smartphone based advanced driver assistance system on distracted driving behavior: a simulator study. Comput. Hum. Behav. 83, 1–7 (2018)CrossRefGoogle Scholar
  7. 7.
    Eftekhari, H.R., Ghatee, M.: A similarity-based neuro-fuzzy modeling for driving behavior recognition applying fusion of smartphone sensors. J. Intell. Transp. Syst. 23, 1–12 (2019)CrossRefGoogle Scholar
  8. 8.
    Pandey, P.S.K., Kulkarni, R.: Traffic sign detection for advanced driver assistance system. In: 2018 International Conference on Advances in Communication and Computing Technology (ICACCT), Sangamner, pp. 182–185 (2018)Google Scholar
  9. 9.
    Yun, D.S., Lee, J., Lee, S., Kwon, O.: Development of the eco-driving and safe-driving components using vehicle information. In: 2012 International Conference on ICT Convergence (ICTC), Jeju Island, pp. 561–562 (2012)Google Scholar
  10. 10.
    Hirschfeld, R.A.: Integration of Vehicle On-Board Diagnostics and Smart Phone Sensors. US20110012720A1, United States Patent and Trademark Office, 20 January 2011Google Scholar
  11. 11.
    Fang, A., Qiu, C., Zhao, L., Jin, Y.: Driver risk assessment using traffic violation and accident data by machine learning approaches. In: 2018 3rd IEEE International Conference on Intelligent Transportation Engineering (ICITE), Singapore, pp. 291–295 (2018)Google Scholar
  12. 12.
    Suzuki, H., Marumo, Y.: A new approach to green light optimal speed advisory (GLOSA) systems and its limitations in traffic flows. In: Proceedings of the 1st International Conference on Human Systems Engineering and Design (IHSED2018): Future Trends and Applications, 25–27 October 2018. CHU-Université de Reims Champagne-Ardenne, France, pp. 776–782 (2019)Google Scholar
  13. 13.
    Biswas, S., Tatchikou, R., Dion, F.: Vehicle-to-vehicle wireless communication protocols for enhancing highway traffic safety. IEEE Commun. Mag. 44(1), 74–82 (2006)CrossRefGoogle Scholar
  14. 14.
    Abdulsattar, H., Mostafizi, A., Siam, M.R.K., Wang, H.: Measuring the impacts of connected vehicles on travel time reliability in a work zone environment: an agent-based approach. In: Journal of Intelligent Transportation Systems Technology Planning and Operations (2019)Google Scholar
  15. 15.
    Kulla, E., Jiang, N., Spaho, E., Nishihara, N.: A survey on platooning techniques in VANETs. In: Complex, Intelligent, and Software Intensive Systems, pp. 650–659. Springer (2019)Google Scholar
  16. 16.
    Mahmood, A., Butler, B., Sheng, Q.Z., Zhang, W.E., Jennings, B.: Need of ambient intelligence for next-generation connected and autonomous vehicles. In: Guide to Ambient Intelligence in the IoT Environment, pp. 133–151. Springer (2019)Google Scholar
  17. 17.
    Aziz, Z., Nandi, A. Microsoft Technology Licensing LLC. Digital assistant for vehicle related activities. US10169794B2, United States Patent and Trademark Office, 01 January 2019Google Scholar
  18. 18.
    Wolverton, M.J., Mark, W.S., Bratt, H., Bercow, D.A.: SRI International. Vehicle personal assistant. US20140136187A1, United States Patent and Trademark Office, 19 January 2012Google Scholar
  19. 19.
    Engelbrecht, J., Booysen, M.J.(Thinus), van Rooyen, G.-J., Bruwer, F.J.: A survey of smartphone-based sensing in vehicles for ITS applications. IET Intell. Transp. Syst. 9, 1–22 (2015)Google Scholar
  20. 20.
    Singh, G., Bansal, D., Sofat, S.: A smartphone-based technique to monitor driving behavior using DTW and crowdsensing. Pervasive Mob. Comput. 40(9), 56–70 (2017)CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.SPIIRASSaint PetersburgRussia
  2. 2.ITMO UniversitySaint PetersburgRussia

Personalised recommendations