No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
The Detection of Visual Distraction using Vehicle and Driver-Based Sensors
... The original sample is randomly partitioned into four equal sized subsamples across the different secondary tasks: For each HSMM1~3, average F-values were respectively 0.91, 0.92, and 0.94, as the harmonic mean of the probability of correctly labeling the detection (recall) and probability that a positive prediction is correct (precision), and the balanced trade-off between recall and precision. Similar to the work on driver state prediction (distracted or not) by [9] Schwarz et al. (2016), the proposed model in this study presents comparably good performance. Average performance of the model was respectively 87%, 85%, and 91%. ...
... The original sample is randomly partitioned into four equal sized subsamples across the different secondary tasks: For each HSMM1~3, average F-values were respectively 0.91, 0.92, and 0.94, as the harmonic mean of the probability of correctly labeling the detection (recall) and probability that a positive prediction is correct (precision), and the balanced trade-off between recall and precision. Similar to the work on driver state prediction (distracted or not) by [9] Schwarz et al. (2016), the proposed model in this study presents comparably good performance. Average performance of the model was respectively 87%, 85%, and 91%. ...
To better understand the effects of distracted driving on crash causation, forward roadway glance durations need to be carefully examined. Secondary tasks that impose high cognitive load lead to spillover effects that are moderated by the duration of the forward roadway glance within an alternation sequence involving both, in-vehicle and on-road glances. Spillover effects diminish the hazard anticipation ability of drivers. When alternating glances in a time series, the probability of detecting a spillover is invisible and the hidden state depends on the amount of time that has elapsed since the secondary task was initiated in the current state which is in contrast with the hidden Markov theory, where there is a constant probability of changing state given spillover detection in the state up to that time. No research estimates the probability of spillover detection in a time series with an explicit glance duration. In the current effort, we apply a semi-hidden Markov model where secondary task severity is used as an observation to infer hidden state and relax the assumption of constant state duration. Based on the reliable accuracy of the task itself, and the proposed model, different sequences of secondary task during various time window were tested for spillover detection. With a threshold of 50%, different forward roadway glance durations are required in each sequence associated with different types of secondary tasks.
... as driver assist systems, alert and guide the driver when attentional reorientation is deemed necessary, and/or 10 provide post-drive feedback to the driver (Kim, Chun, & Dey, 2015;Lee et al., 2013;Schwarz, Brown, Lee, 11 Gaspar, & Kang, 2016;Smith, Witt, Bakowski, Leblanc, & Lee, 2009). While most prior work in distraction 12 estimation is based on linear or logistic modeling approaches, more recent work has investigated machine 13 learning approaches to facilitate modeling highly non-linear behavior that can be more sensitive to some types 14 of distraction (e.g., Schwarz et al., 2016;Masood et al., 2018). 15 ...
... Existing patterns of data associated with different types of distracted driving can be used to 17 develop an algorithm capable of predicting future, unlabeled patterns (Dong, Hu, Uchimura, & Murayama, 18 2011;Kotsiantis, 2007). The large body of research on driving distraction has documented how different types 19 of distraction can affect, vehicle control input ( Dingus et al., 2016;Drews, Yazdani, Godfrey, Cooper, & 20 Strayer, 2009;Engström et al., 2017;Feng et al., 2017;Horrey & Wickens, 2006;Strayer, Drews, & Johnston, 21 2003), driver head posture and eye gaze ( Lee et al., 2013;Schwarz et al., 2016;Tippey, Sivaraj, & Ferris, 22 2017), and physiological indicators of arousal in a driver's sympathetic nervous system, such as heart rate 23 measures, galvanic skin response, or perinasal perspiration (Collet, Guillot, & Petit, 2010;Healey & Picard, 24 2005;Kim et al., 2015;Mehler, Reimer, Coughlin, & Dusek, 2009;Pavlidis et al., 2016;Reimer, Mehler,Coughlin, Roy, & Dusek, 2011). These observable measures can therefore be consulted to provide varying 26 amounts of evidence of a distracted driver. ...
Objective. The objective of this study was to analyze a set of driver performance and physiological data using advanced machine learning approaches, including feature generation, to determine the best-performing algorithms for detecting driver distraction and predicting the source of distraction.
Background. Distracted driving is a causal factor in many vehicle crashes, often resulting in injuries and deaths. As mobile devices and in-vehicle information systems become more prevalent, the ability to detect and mitigate driver distraction becomes more important.
Method. This study trained twenty-one algorithms to identify when drivers were distracted by secondary cognitive and texting tasks. The algorithms included physiological and driving behavioral input, processed with a comprehensive feature generation package, Time Series Feature Extraction based on Scalable Hypothesis tests.
Results. Results showed that a Random Forest algorithm, trained using only driving behavior measures and excluding driver physiological data, was the highest-performing algorithm for accurately classifying driver distraction. The most important input measures identified were lane offset, speed and, steering while the most important feature types were standard deviation, quantiles, and non-linear transforms.
Conclusion. This work suggests that distraction detection algorithms may be improved by considering ensemble machine-learning algorithms that are trained with driving behavior measures and non-standard features. Additionally, the study presents several new indicators of distraction derived from speed and steering measures.
Application. Future development of distraction mitigation systems should focus on driver behavior-based algorithms that use complex feature generation techniques.
... The driver's eye movement and vehicle performance were integrated as a real-time cognitive DD attribute [4,30,31] and the SVM algorithm was applied in these studies. Driving performance and head movement tracking were integrated for the DD detection with random forest model and HMM [32]. In [33], different machine learning methods, SVM, conventional recurrent NN, and long or short-term memory recurrent NN, using the same attributes were compared for continuously driver's state prediction. ...
... For the same reason, the methods with different attribute combinations (e.g. behavioral, psychological, and subjective) [30][31][32][33][34][35] are not feasible in the real world implementation. ...
In addition to vehicle control, drivers often perform secondary tasks that impede driving. Reduction of driver distraction is an important challenge for the safety of intelligent transportation systems. In this paper, a methodology for the detection and evaluation of driver distraction while performing secondary tasks is described and an appropriate hardware and a software environment is offered and studied. The system includes a model of normal driving, a subsystem for measuring the errors from the secondary tasks, and a module for total distraction evaluation. A new machine learning algorithm defines driver performance in lane keeping and speed maintenance on a specific road segment. To recognize the errors, a method is proposed, which compares normal driving parameters with ones obtained while conducting a secondary task. To evaluate distraction, an effective fuzzy logic algorithm is used. To verify the proposed approach, a case study with driver-in-the-loop experiments was carried out, in which participants performed the secondary task, namely chatting on a cell phone. The results presented in this research confirm its capability to detect and to precisely measure a level of abnormal driver performance.
... Another option is to use some other distraction detection algorithm as ground truth. The original AttenD algorithm has actually been used as ground truth of distraction [57] as well as for attention management [58]. This approach does not really solve the ground truth issue though, and it hampers further developments in the field, since it is impossible to be better than the "ground truth". ...
This paper presents initial work on a context-dependent driver distraction detection algorithm called AttenD2.0, which extends the original AttenD algorithm with elements from the Minimum Required Attention (MiRA) theory. Central to the original AttenD algorithm is a time buffer which keeps track of how often and for how long the driver looks away from the forward roadway. When the driver looks away the buffer is depleted and when looking back the buffer fills up. If the buffer runs empty the driver is classified as distracted. AttenD2.0 extends this concept by adding multiple buffers, thus integrating situation dependence and visual time-sharing behaviour in a transparent manner. Also, the increment and decrement of the buffers are now controlled by both static requirements (e.g. the presence of an on-ramp increases the need to monitor the sides and the mirrors) as well as dynamic requirements (e.g., reduced speed lowers the need to monitor the speedometer). The algorithm description is generic, but a real-time implementation with concrete values for different parameters is showcased in a driving simulator experiment with 16 bus drivers, where AttenD2.0 was used to ensure that drivers are attentive before taking back control after an automated bus stop docking and depot procedure. The scalability of AttenD2.0 relative to available data sources and the level of vehicle automation is demonstrated. Future work includes expanding the concept to real-world environments by automatically integrating situational information from the vehicles environmental sensing and from digital maps.
... In [13], a different approach was implemented for detecting visual distractions. Unlike the other studies mentioned beforehand, vehicle-based sensors were developed to detect distractions. ...
... While the statistics present a case that driver error is a contributor to the problem of SUA due to revealed demographic differences, it is not the case that age and/or gender can simply be used as predictors for the occurrence of SUA. Driver behavior is quite individual and depends on many factors, yet it can be predictive of driver impairment and the adverse events that follow McDonald et al., , 2012Schwarz et al., 2015Schwarz et al., , 2016. If characteristics of parking, for example, can be identified and customized to the individual (e.g., in the form of parking signatures), then unusual patterns might be detected and used to detect increased risk of serious events such as SUA. ...
Pedal misapplications by drivers have received attention as being an underlying factor for the phenomenon known as sudden unintended acceleration (SUA) in vehicles. This research investigates behaviors during a common task for drivers, namely residential parking. Parking has been identified as a maneuver that is often linked with SUA mishaps. Using driving trajectories data from a set of four couples collected as part of a naturalistic driving study, we investigate whether consistent behaviors can be detected when parking at home from a geospatial perspective, i.e., whether deceleration and braking occur in a characteristic way at the end of a driving trajectory, and whether these behaviors vary when the geospatial context of parking changes. An ontology-based approach is used to frame the key behaviors of the naturalistic driving, and big data techniques are applied to extract parking-specific behaviors from driving trajectories. Results show that individuals showed relatively consistent parking behaviors under the same geospatial context and the standard deviation of the deceleration threshold has a larger discrepancy between couples parking at different residences than within couples where parking occurs at the same place.
... The modeling of driver impairment has been conducted using not only driver-based sensors (eye tracking), but also vehicle-based ones (steering, pedals). This approach has been used at the National Advanced Driving Simulator (NADS) by applying machine learning techniques to identify driving patterns unique to alcohol (Lee et al., 2010), drowsiness (Brown, Lee, Schwarz, Fiorentino, & McDonald, 2014) and distraction (Schwarz, Brown, & Gaspar, 2016). ...
The National Advanced Driving Simulator is a high-fidelity motion-base simulator owned by the National Highway Transportation Safety Administration and managed and operated by the University of Iowa. Its 25-year history has intersected with some of the most significant developments in automotive history, such as advanced driver assistance systems like stability control and collision warning systems, and highly automated vehicles. The simulator is an application of immersive virtual reality that uses multiprojection instead of head-mounted displays. A large-excursion motion system provides realistic acceleration and rotation cues to the driver. Due to its level of immersion and realism, drivers respond to events in the simulator the same way they would in their own vehicle. We document the history and technology behind this national facility.
The Vehicle Active Safety Technology has attracted considerable attention in the development of Vehicle Intelligent Technology. It not only performs a significant role in vehicle safety but also changes driving behaviors. Because of the advantages of Driving Simulator, numerous researchers have made efforts to popularize Driving Simulator Technology in the field of vehicle safety. The driving simulators are used for making validation test in such fields which are cover all the research fields in vehicle safety technology, including Electronic Stability Control, Active Front Steering, Driver Monitoring Technology, and Collision Avoidance Technology. The functions of Driving Simulator vary from algorithm verification to system fusion. While Driving Simulators on conducting a few validations have been extensively studied in vehicle safety tests, there were few researchers in summarizing the applications of Driving Simulator. To make it clear the formation and development of Driving Simulator Technology, this paper presents the state-of-the-art applications of developing driving simulator in the vehicle active safety system. Besides, it also discusses the relevant information on what a typical Driving Simulator is with an example of the driving simulator of Chang’an University.
This paper quantifies the safety benefits of a proposed near real-time traffic control system for highway–rail grade crossing (HRGC) utilizing emerging safety technologies in connected and autonomous vehicles (CAV). The connected-vehicle technologies that have applications at a railroad crossing include vehicle-based technologies (railroad crossing violation warning, automated or semi-automated braking system, drowsiness/distracted driver alert) and technologies that require cooperation from the railroad industry (advanced warnings to trains about an occupied crossing). This paper provides a methodology to quantify the reduction in crashes as safety technologies become prevalent. It first identifies crash characteristics that enable the classification of crashes as preventable crashes or not. This classification is used to train machine learning models to estimate the likelihood of a potential crash being preventable. The machine learning model is used along with the zero inflated negative binomial with empirical Bayes system (ZINEBS) model to estimate the expected accident count at a crossing when a percentage of vehicles in the traffic stream is CAV. This paper presents case studies for three crossings to show the reduction in crashes with the increase in the percentage of connected vehicles in the traffic stream. It also presents the general trend in the reduction expected by analyzing 50 crossings of each warning device type.
Factors related to distraction have been one of the major causes of accidents related to driving. This paper presents a model that detects distraction cues of a Grab driver that mainly focuses on visual distraction and continuous communication indicated by mouth movements. The data gathered was in the form of a video footage obtained from each transit recorded. Each video was processed frame by frame to extract necessary features for detecting distraction cues through OpenFace. Binary classification was used to assess whether the driver is distracted or not. To determine the predictive power of the model, K-fold cross validation method was used. Since Grab has been widely used in the Philippines, the findings would allow for interesting knowledge regarding the behavior that these drivers exhibited in varying situations with regard to distraction.
Performance of in-vehicle, secondary tasks requires a driver to alternate his or her glances between the inside of the vehicle and the forward roadway. While previous research has shown that thresholds of off-road and durations are critical to latent hazard detection, there has been no research to predict the probability of hazard detection in a time series considering all possible forward glance durations. To determine the minimum forward roadway duration to adequately anticipate hazards, 45 drivers were asked to navigate eight unique virtual scenarios on a driving simulator while alternating their glances inside (2s) and outside (1s, 2s, 3s, or 4s) the vehicle in a uniform sequence (consistent on-road and off-road durations). A microbenchmark approach based on Hidden Markov models is explored to infer the transition probability of hazard detection that changes dynamically between the detection and nondetection stages. The model is cross validated and demonstrated to be accurate and robust. Three characteristics of total sequence time were tested in the model. Using the ground truth transition probability from fixed duration, the probability of hazard detection with variable duration within an alternation sequence was computed. Across nonuniform alternation sequences, different permutations of sequences show that a short time series of alternation (10s) of glance behavior is sufficient for hazard detection (greater than 50.0%). Appropriate countermeasures can be developed to increase a driver's forward glance duration whenever the detection probability is predicted low.
The “100-Car Naturalistic Driving Study” is a three-phased effort designed to accomplish three objectives: Phase I, Conduct Test Planning Activities; Phase II, Conduct a Field Test; and Phase III, Prepare for Large-Scale Field Data Collection Effort. This report documents the efforts of Phase II. Project sponsors are the National Highway Traffic Safety Administration (NHTSA) and the Virginia Department of Transportation (VDOT).
The 100-Car Naturalistic Driving Study is the first instrumented-vehicle study undertaken with the primary purpose of collecting large-scale, naturalistic driving data. Drivers were given no special instructions, no experimenter was present, and the data collection instrumentation was unobtrusive. In addition, 78 of 100 vehicles were privately owned. The resulting database contains many extreme cases of driving behavior and
performance, including severe drowsiness, impairment, judgment error, risk taking, willingness to engage in secondary tasks, aggressive driving, and traffic violations. The data set includes approximately 2,000,000 vehicle miles, almost 43,000 hours of data, 241 primary and secondary drivers, 12 to 13 months of data collection for each vehicle, and data from a highly capable instrumentation system including 5 channels of video and many vehicle state and kinematic sensors. From the data, an “event” database was created, similar in classification structure to an epidemiological crash database, but with video and electronic driver and vehicle performance data. The events are crashes, near-crashes, and other “incidents.” Data is classified by pre-event maneuver, precipitating factor, event type, contributing factors, associative factors, and the avoidance maneuver. Parameters such as vehicle speed, vehicle headway, time-to-collision, and driver reaction time are also recorded.
The current project specified ten objectives or goals that would be addressed through the initial analysis of the event database. This report addresses the first 9 of these goals, which include analyses of rear-end events, lane change events, the role of inattention, and the relationship between levels of severity. Goal 10 is a separate report and addresses the implications for a larger-scale data collection effort.
This work was sponsored by the second Strategic Highway Research Program (SHRP 2), which is administered by the Transportation Research Board of the National Academies. This project was managed by Ken Campbell, Chief Program Officer for SHRP 2 Safety , and Jim Hedlund, SHRP 2 Safety Coordinator . The research reported on herein was performed by the main contractor SAFER Vehicle and Traffic Safety Centre at Chalmers, Gothenburg, Sweden. SAFER is a joint research unit where 25 partners from the Swedish automotive industry, academia and authorities cooperate to make a center of excellence within the field of vehicle and traffic safety (see www.chalmers.se/safer ). The host and legal entity SAFER is Chalmers University of Technology. Principle Investigator Tr ent Victor is Adjunct Professor at Chalmers and worked on the project as borrowed personnel to Chalmers but his main employer is Volvo Cars. The other authors of this report are Co - PI Marco Dozza, Jonas Bärgman, and Christian - Nils Boda of Chalmers Universi ty of Technology (as a SAFER partner) ; Johan Engström and Gustav Markkula of Volvo Group Trucks Technology (as a SAFER partner) ; John D. Lee of University of Wisconsin - Madison (as a consultant to SAFER); and Carol Flannagan of University of Michigan Transp ortation Research Institute (UMTRI) (as a consultant to SAFER). The authors acknowledge the contributions to this research from Ines Heinig, Vera Lisovskaja, Olle Nerman, Holger Rootzén, Dmitrii Zholud, Helena Gellerman , Leyla Vujić, Martin Rensfeldt, Stefan Venbrant, Akhil Krishnan, Bharat Mohan Redrouthu, Daniel Nilsson of Chalmers; Mikael Ljung - Aust of Volvo Cars; Erwin Boer; Christer Ahlström and Omar Bagdadi of VTI.
Despite persistent efforts at the local, state, and federal levels, alcohol-impaired driving crashes still account for 31% of all traffic fatalities. The proportion of fatally injured drivers with blood alcohol concentrations (BAC) greater than or equal to 0.08% has remained at 31-32% for the past ten years. Vehicle-based countermeasures have the potential to address this problem and save thousands of lives each year. Many of these vehicle-based countermeasures depend on developing an algorithm that uses driver performance to assess impairment. The National Advanced Driving Simulator (NADS) was used to collect data needed to develop an algorithm for detecting alcohol impairment. Data collection involved 108 drivers from three age groups (21-34, 38-51, and 55-68 years of age) driving on three types of roadways (urban, freeway, and rural) at three levels of alcohol concentration (0.00%, 0.05%, and 0.10% BAC). The scenarios used for this data collection were selected so that they were both representative of alcohol-impaired driving and sensitive to alcohol impairment. The data from these scenarios supported the development of three algorithms. One algorithm used logistic regression and standard speed and lane-keeping measures; a second used decision trees and a broad range of driving metrics that are grounded in cues NHTSA has suggested police officers use to identify alcohol-impaired drivers; a third used a support vector machines. The results demonstrate the feasibility of a vehicle-based system to detect alcohol impairment based on driver behavior. The algorithms differentiate between drivers with BAC levels at and above and below 0.08%BAC with an accuracy of 73 to 86%, comparable to the standardized field sobriety test. This accuracy can be achieved with approximately eight minutes of driving performance data. Differences between drivers and between roadway situations have a large influence on algorithm performance, which suggests the algorithms should be tailored to drivers and to road situations.
Many statistical methods rely on an underlying mathematical model of probability based on a simple approximation, one that is simultaneously well-known and yet frequently misunderstood. The Normal approximation to the Binomial distribution underpins a range of statistical tests and methods, including the calculation of accurate confidence intervals, performing goodness of fit and contingency tests, line- and model-fitting, and computational methods based upon these. A common mistake is in assuming that, since the probable distribution of error about the “true value” in the population is approximately Normally distributed, the same can be said for the error about an observation.This paper is divided into two parts: fundamentals and evaluation. First, we examine the estimation of confidence intervals using three initial approaches: the “Wald” (Normal) interval, the Wilson score interval and the “exact” Clopper-Pearson Binomial interval. Whereas the first two can be calculated directly from formulae, the Binomial interval must be approximated towards by computational search, and is computationally expensive. However this interval provides the most precise significance test, and therefore will form the baseline for our later evaluations. We also consider two further refinements: employing log-likelihood in intervals (also requiring search) and the effect of adding a continuity correction.Second, we evaluate each approach in three test paradigms. These are the single proportion interval or 2 × 1 goodness of fit test, and two variations on the common 2 × 2 contingency test. We evaluate the performance of each approach by a “practitioner strategy”. Since standard advice is to fall back to “exact” Binomial tests in conditions when approximations are expected to fail, we report the proportion of instances where one test obtains a significant result when the equivalent exact test does not, and vice versa, across an exhaustive set of possible values.We demonstrate that optimal methods are based on continuity-corrected versions of the Wilson interval or Yates’ test, and that commonly-held beliefs about weaknesses of tests are misleading. Log-likelihood, often proposed as an improvement on , performs disappointingly. Finally we note that at this level of precision we may distinguish two types of 2 2 test according to whether the independent variable partitions data into independent populations, and we make practical recommendations for their use.
We present a method to monitor driver distraction based on a stereo camera to estimate the face pose and gaze of a driver in real time. A coarse eye direction is composed of face pose estimation to obtain the gaze and driver's fixation area in the scene, which is a parameter that gives much information about the distraction pattern of the driver. The system does not require any subject-specific calibration; it is robust to fast and wide head rotations and works under low-lighting conditions. The system provides some consistent statistics, which help psychologists to assess the driver distraction patterns under influence of different in-vehicle information systems (IVISs). These statistics are objective, as the drivers are not required to report their own distraction states. The proposed gaze fixation system has been tested on a set of challenging driving experiments directed by a team of psychologists in a naturalistic driving simulator. This simulator mimics conditions present in real driving, including weather changes, maneuvering, and distractions due to IVISs. Professional drivers participated in the tests.
Driver cognitive distraction (e.g., hand-free cell phone conversation) can lead to unapparent, but detrimental, impairment to driving safety. Detecting cognitive distraction represents an important function for driver distraction mitigation systems. We developed a layered algorithm that integrated two data mining methods—Dynamic Bayesian Network (DBN) and supervised clustering—to detect cognitive distraction using eye movement and driving performance measures. In this study, the algorithm was trained and tested with the data collected in a simulator-based study, where drivers drove either with or without an auditory secondary task. We calculated 19 distraction indicators and defined cognitive distraction using the experimental condition (i.e., “distraction” as in the drives with the secondary task, and “no distraction” as in the drives without the secondary task). We compared the layered algorithm with previously developed DBN and Support Vector Machine (SVM) algorithms. The results showed that the layered algorithm achieved comparable prediction performance as the two alternatives. Nonetheless, the layered algorithm shortened training and prediction time compared to the original DBN because supervised clustering improved computational efficiency by reducing the number of inputs for DBNs. Moreover, the supervised clustering of the layered algorithm revealed rich information on the relationship between driver cognitive state and performance. This study demonstrates that the layered algorithm can capitalize on the best attributes of component data mining methods and can identify human cognitive state efficiently. The study also shows the value in considering the supervised clustering method as an approach to feature reduction in data mining applications.
Driver drowsiness is among the leading causal factors in traffic accidents occurring worldwide. This paper describes a method to monitor driver safety by analyzing information related to fatigue using two distinct methods: eye movement monitoring and bio-signal processing. A monitoring system is designed in Android-based smartphone where it receives sensory data via wireless sensor network and further processes the data to indicate the current driving aptitude of the driver. It is critical that several sensors are integrated and synchronized for a more realistic evaluation of the driver behavior. The sensors applied include a video sensor to capture the driver image and a bio-signal sensor to gather the driver photoplethysmograph signal. A dynamic Bayesian network framework is used for the driver fatigue evaluation. A warning alarm is sounded if driver fatigue is believed to reach a defined threshold. The manifold testing of the system demonstrates the practical use of multiple features, particularly with discrete methods, and their fusion enables a more authentic and ample fatigue detection.
The effects of eight in-vehicle tasks on driver distraction were measured in a large, moving-base driving simulator. Forty-eight adults, ranging in age from 35 to 66, and 15 teenagers participated in the simulated drive. Hand-held and hands-free versions of phone dialing, voicemail retrieval, and incoming calls represented six of the eight tasks. Manual radio tuning and climate control adjustment were also included to allow comparison with tasks that have traditionally been present in vehicles. During the drive the participants were asked to respond to sudden movements in surrounding traffic. The driver's ability to detect these sudden movements or events changed with the nature of the in-vehicle tasks that were being performed. Driving performance measures such as lane violations and heading error were also computed. The performance of the adult group was compared with the performance of the teenage drivers. Compared with the adults, the teens were found to choose unsafe following distances, have poor vehicle control skills, and be more prone to distraction from hand-held phone tasks.
Gaze tracking is crucial for studying driver's attention, detecting fatigue, and improving driver assistance systems, but it is difficult in natural driving environments due to nonuniform and highly variable illumination and large head movements. Traditional calibrations that require subjects to follow calibrators are very cumbersome to be implemented in daily driving situations. A new automatic calibration method, based on a single camera for determining the head orientation and which utilizes the side mirrors, the rear-view mirror, the instrument board, and different zones in the windshield as calibration points, is presented in this paper. Supported by a self-learning algorithm, the system tracks the head and categorizes the head pose in 12 gaze zones based on facial features. The particle filter is used to estimate the head pose to obtain an accurate gaze zone by updating the calibration parameters. Experimental results show that, after several hours of driving, the automatic calibration method without driver's corporation can achieve the same accuracy as a manual calibration method. The mean error of estimated eye gazes was less than 5° in day and night driving.
Most driver-monitoring systems have attempted to detect either driver
drowsiness or distraction, although both factors should be considered
for accident prevention. Therefore, we propose a new driver-monitoring
method considering both factors. We make the following contributions.
First, if the driver is looking ahead, drowsiness detection is
performed; otherwise, distraction detection is performed. Thus, the
computational cost and eye-detection error can be reduced. Second, we
propose a new eye-detection algorithm that combines adaptive boosting,
adaptive template matching, and blob detection with eye validation,
thereby reducing the eye-detection error and processing time
significantly, which is hardly achievable using a single method. Third,
to enhance eye-detection accuracy, eye validation is applied after
initial eye detection, using a support vector machine based on
appearance features obtained by principal component analysis (PCA) and
linear discriminant analysis (LDA). Fourth, we propose a novel eye
state-detection algorithm that combines appearance features obtained
using PCA and LDA, with statistical features such as the sparseness and
kurtosis of the histogram from the horizontal edge image of the eye.
Experimental results showed that the detection accuracies of the eye
region and eye states were 99 and 97%, respectively. Both driver
drowsiness and distraction were detected with a success rate of 98%.
This chapter discusses our research efforts in tracking driver distraction using multimodal features. A car equipped with various sensors is used to collect a database with real driving conditions. During the recording, the drivers were asked to perform common secondary tasks such as operating a cell phone, talking to another passenger and changing the radio stations. We analyzed the differences observed across multimodal features when the driver is engaged in these secondary tasks. The study considers features extracted from the controller area network-bus (CAN-Bus), a frontal camera facing the driver and a microphone. These features are used to predict the distraction level of the drivers. The output of the proposed regression model has high correlation with human subjective evaluations (ρ=0.728), which validates our approach.
To assess the sensitivity of two physiological measures for discriminating between levels of cognitive demand under driving conditions across different age groups.
Previous driving research presents a mixed picture concerning the sensitivity of physiological measures for differentiating tasks with presumed differences in mental workload.
A total of 108 relatively healthy drivers balanced by gender and across three age groups (20-29, 40-49, 60-69) engaged in three difficulty levels of an auditory presentation-verbal response working memory task.
Heart rate and skin conductance level (SCL) both increased in a statistically significant fashion with each incremental increase in cognitive demand, whereas driving performance measures did not provide incremental discrimination. SCL was lower in the 40s and 60s age groups; however, the pattern of incremental increase with higher demand was consistent for heart rate and SCL across all age groups. Although each measure was quite sensitive at the group level, considering both SCL and heart rate improved detection of periods of heightened cognitive demand at the individual level.
The data provide clear evidence that two basic physiological measures can be utilized under field conditions to differentiate multiple levels of objectively defined changes in cognitive demand. Methodological considerations, including task engagement, may account for some of the inconsistencies in previous research.
These findings increase the confidence with which these measures may be applied to assess relative differences in mental workload when developing and optimizing human machine interface (HMI) designs and in exploring their potential role in advanced workload detection and augmented cognition systems.
Monitoring driver fatigue, inattention, and lack of sleep is very important in preventing motor vehicles accidents. A visual system for automatic driver vigilance has to address two fundamental problems. First of all, it has to analyze the sequence of images and detect if the driver has his eyes open or closed, and then it has to evaluate the temporal occurrence of eyes open to estimate the driver's visual attention level. In this paper we propose a visual approach that solves both problems. A neural classifier is applied to recognize the eyes in the image, selecting two candidate regions that might contain the eyes by using iris geometrical information and symmetry. The novelty of this work is that the algorithm works on complex images without constraints on the background, skin color segmentation and so on. Several experiments were carried out on images of subjects with different eye colors, some of them wearing glasses, in different light conditions. Tests show robustness with respect to situations such as eyes partially occluded, head rotation and so on. In particular, when applied to images where people have eyes closed the proposed algorithm correctly reveals the absence of eyes. Next, the analysis of the eye occurrence in image sequences is carried out with a probabilistic model to recognize anomalous behaviors such as driver inattention or sleepiness. Image sequences acquired in the laboratory and while people were driving a car were used to test the driver behavior analysis and demonstrate the effectiveness of the whole approach.
The objective of this research was to assess the driver visual attentional demand requirements of an operational in-car navigation system. Thirty-two driver subjects ranging in age from 18 to 73 drove a specially instrumented vehicle on various types of public roadways with varying traffic conditions. Drivers performed both specific navigation system tasks and conventional tasks using dashboard instrumentation. Comparisons were then made between the two types of tasks. Results show that the demand of most of the navigation tasks was comparable to that of one or more conventional tasks. Modifying the remaining navigation tasks to make information more readily available would reduce their demand.
The underlying aim of HASTE, an EU FP5 project, is the development of a valid, cost-effective and reliable assessment protocol to evaluate the potential distraction of an in-vehicle information system on driving performance. As part of this development, the current study was performed to examine the systematic relationship between primary and secondary task complexity for a specific task modality in a particular driving environment. Two fundamentally distinct secondary tasks (or surrogate in-vehicle information systems, sIVIS) were developed: a visual search task, designed such that it only required visual processing/demand and an auditory continuous memory task, intended to cognitively load drivers without any visual stimulus. A high fidelity, fixed-base driving simulator was used to test 48 participants on a car following task. Virtual traffic scenarios varied in driving demand. Drivers compensated for both types of sIVIS by reducing their speed (this result was more prominent during interaction with the visual task). However, they seemed incapable of fully prioritising the primary driving task over either the visual or cognitive secondary tasks as an increase in sIVIS demand was associated with a reduction in driving performance: drivers showed reduced anticipation of braking requirements and shorter time-to-collision. These results are of potential interest to designers of in-vehicle systems.
This paper describes the R package mhsmm which implements estimation and prediction methods for hidden Markov and semi-Markov models for multiple observation sequences. Such techniques are of interest when observed data is thought to be dependent on some unobserved (or hidden) state. Hidden Markov models only allow a geometrically distributed sojourn time in a given state, while hidden semi-Markov models extend this by allowing an arbitrary sojourn distribution. We demonstrate the software with simulation examples and an application involving the modelling of the ovarian cycle of dairy cows.
Receiver operating characteristic (ROC) curves are useful tools to evaluate classifiers in biomedical and bioinformatics applications. However, conclusions are often reached through inconsistent use or insufficient statistical analysis. To support researchers in their ROC curves analysis we developed pROC, a package for R and S+ that contains a set of tools displaying, analyzing, smoothing and comparing ROC curves in a user-friendly, object-oriented and flexible interface.
With data previously imported into the R or S+ environment, the pROC package builds ROC curves and includes functions for computing confidence intervals, statistical tests for comparing total or partial area under the curve or the operating points of different classifiers, and methods for smoothing ROC curves. Intermediary and final results are visualised in user-friendly interfaces. A case study based on published clinical and biomarker data shows how to perform a typical ROC analysis with pROC.
pROC is a package for R and S+ specifically dedicated to ROC analysis. It proposes multiple statistical tests to compare ROC curves, and in particular partial areas under the curve, allowing proper ROC interpretation. pROC is available in two versions: in the R programming language or with a graphical user interface in the S+ statistical software. It is accessible at http://expasy.org/tools/pROC/ under the GNU General Public License. It is also distributed through the CRAN and CSAN public repositories, facilitating its installation.
A new metaheuristic optimisation algorithm, called Cuckoo Search (CS), was developed recently by Yang and Deb (2009). This paper presents a more extensive comparison study using some standard test functions and newly designed stochastic test functions. We then apply the CS algorithm to solve engineering design optimisation problems, including the design of springs and welded beam structures. The optimal solutions obtained by CS are far better than the best solutions obtained by an efficient particle swarm optimiser. We will discuss the unique search features used in CS and the implications for further research. Comment: 14 pages; Yang, X.-S., and Deb, S. (2010), Engineering Optimisation by Cuckoo Search
The caret package, short for classification and regression training, contains numerous tools for developing predictive models using the rich set of models available in R. The package focuses on simplifying model training and tuning across a wide variety of modeling techniques. It also includes methods for pre-processing training data, calculating variable importance, and model visualizations. An example from computational chemistry is used to illustrate the functionality on a real data set and to benchmark the benefits of parallel processing with several types of models.
This study investigates the impact of multiple in-vehicle information systems on the driver. It was undertaken using a high fidelity driving simulator. The participants experienced, paced and unpaced single tasks, multiple secondary tasks and an equal period of 'normal' driving. Results indicate that the interaction with secondary tasks led to significant compensatory speed reductions. Multiple secondary tasks were shown to have a detrimental affect on vehicle performance with significantly reduced headways and increased brake pressure being found. The drivers reported interaction with the multiple in-vehicle systems to significantly impose more subjective mental workload than either a single secondary task or 'normal driving'. The implications of these findings and the need to integrate and manage complex in-vehicle information systems are discussed.
Drowsy driving is a significant contributor to death and injury crashes on our Nation's highways, accounting for more than 80,000 crashes and 850 fatalities per year. The successful detection of drowsiness is a crucial step in implementing mitigation strategies to reduce the cost to society of drowsy driving. Building upon prior research in detecting impairment from alcohol and distraction, the goal of this research was to determine the extent to which alcohol impairment algorithms could detect drowsiness and distinguish it from alcohol impairment. Data were collected from seventy-two participants during daytime (9 a.m. - 1 p.m.), early night (10 p.m. - 2 a.m.), and late night (2 a.m. - 6 a.m.) sessions to provide data for algorithm testing and refinement. Driving data indicated a complex relationship between driving performance and conditions associated with drowsiness: compared to daytime session, driving performance improved during the early night session, before degrading during the late night session. This non-linear relationship between continuous time awake, subjective assessments of drowsiness and driving performance has the potential to complicate the early detection of drowsiness. Drowsiness, as indicated by unintended lane departures, occurred in all sessions and demonstrated a transient nature. Algorithms based on lane position and steering wheel data, which can be obtained inexpensively, were best at predicting drowsiness related lane departures. Alcohol detection algorithms were not successful in detecting drowsiness but could be retrained to do so. Rather than one algorithm being generalized to detect multiple impairments, these results indicate that specialized algorithms might co-exist and allow one to detect and differentiate alcohol and drowsy-impaired driving. These findings provide a better understanding of the relationship between impairment from alcohol and drowsiness and lay the foundation for detecting and differentiating among impairment from alcohol, drowsiness, fatigue and drugs.
Driver distraction is a contributing factor to many crashes, and a real-time distraction warning system has the potential to mitigate or circumvent many of these crashes. The objective of this chapter is to describe in detail the distraction detection algorithm AttenD and explain the theory underlying different design choices. Future aspects and distraction warning strategies are discussed also. In summary, AttenD is an eye-tracker-based distraction detection algorithm which identifies visual distraction in real time based on single long glances as well as repetitive glances. The core idea of the algorithm is a two-second time buffer which is decremented when the driver looks away from the road and incremented when the driver looks back at the road. If the buffer runs empty, the driver's state is classified as distracted.
Random forests are a combination of tree predictors such that each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest. The generalization error for forests converges a.s. to a limit as the number of trees in the forest becomes large. The generalization error of a forest of tree classifiers depends on the strength of the individual trees in the forest and the correlation between them. Using a random selection of features to split each node yields error rates that compare favorably to Adaboost (Y. Freund & R. Schapire, Machine Learning: Proceedings of the Thirteenth International conference, ***, 148-156), but are more robust with respect to noise. Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the splitting. Internal estimates are also used to measure variable importance. These ideas are also applicable to regression.
Objective:
The aim of this study was to design and evaluate an algorithm for detecting drowsiness-related lane departures by applying a random forest classifier to steering wheel angle data.
Background:
Although algorithms exist to detect and mitigate driver drowsiness, the high rate of false alarms and missed detection of drowsiness represent persistent challenges. Current algorithms use a variety of data sources, definitions of drowsiness, and machine learning approaches to detect drowsiness.
Method:
We develop a new approach for detecting drowsiness-related lane departures using steering wheel angle data that employ an ensemble definition of drowsiness and a random forest algorithm. Data collected from 72 participants driving the National Advanced Driving Simulator are used to train and evaluate the model. The model's performance was assessed relative to a commonly used algorithm, percentage eye closure (PERCLOS).
Results:
The random forest steering algorithm had a higher classification accuracy and area under the receiver operating characteristic curve than PERCLOS and had comparable positive predictive value. The algorithm succeeds at identifying two key scenarios associated with the drowsiness detection task. These two scenarios consist of instances when drivers depart their lane because they fail to modulate their steering behavior according to the demands of the simulated road and instances when drivers correctly modulate their steering behavior according to the demands of the road.
Conclusion:
The random forest steering algorithm is a promising approach to detect driver drowsiness. The algorithm's ties to consequences of drowsy driving suggest that it can be easily paired with mitigation systems.
The ability to classify driver behavior lays the foundation for more advanced driver assistance systems. In particular, improving safety at intersections has been identified as a high priority due to the large number of intersection-related fatalities. This paper focuses on developing algorithms for estimating driver behavior at road intersections and validating them on real traffic data. It introduces two classes of algorithms that can classify drivers as compliant or violating. They are based on (1) support vector machines and (2) hidden Markov models, which are two very popular machine learning approaches that have been used successfully for classification in multiple disciplines. However, existing work has not explored the benefits of applying these techniques to the problem of driver behavior classification at intersections. The developed algorithms are successfully validated using naturalistic intersection data collected in Christiansburg, VA, through the U.S. Department of Transportation Cooperative Intersection Collision Avoidance System for Violations initiative. Their performances are also compared with those of three traditional methods, and the results show significant improvements with the new algorithms.
Random forests are a combination of tree predictors such that each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest. The generalization error for forests converges a.s. to a limit as the number of trees in the forest becomes large. The generalization error of a forest of tree classifiers depends on the strength of the individual trees in the forest and the correlation between them. Using a random selection of features to split each node yields error rates that compare favorably to Adaboost (Y. Freund & R. Schapire, Machine Learning: Proceedings of the Thirteenth International conference, ***, 148–156), but are more robust with respect to noise. Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the splitting. Internal estimates are also used to measure variable importance. These ideas are also applicable to regression.
Driver distraction and driver inattention are frequently recognized as leading causes of crashes and incidents. Despite this fact, there are few methods available for the automatic detection of driver distraction. Eye tracking has come forward as the most promising detection technology, but the technique suffers from quality issues when used in the field over an extended period of time. Eye-tracking data acquired in the field clearly differs from what is acquired in a laboratory setting or a driving simulator, and algorithms that have been developed in these settings are often unable to operate on noisy field data. The aim of this paper is to develop algorithms for quality handling and signal enhancement of naturalistic eye- and head-tracking data within the setting of visual driver distraction. In particular, practical issues are highlighted. Developed algorithms are evaluated on large-scale field operational test data acquired in the Sweden–Michigan Field Operational Test (SeMiFOT) project, including data from 44 unique drivers and more than 10 000 trips from 13 eye-tracker-equipped vehicles. Results indicate that, by applying advanced data-processing methods, sensitivity and specificity of eyes-off-road glance detection can be increased by about 10%. In conclusion, postenhancement and quality handling is critical when analyzing large databases with naturalistic eye-tracking data. The presented algorithms provide the first holistic approach to accomplish this task.
We present novel algorithms for predictive tracking of user position and orientation based on double exponential smoothing. These algorithms, when compared against Kalman and extended Kalman filter-based predictors with derivative free measurement models, run approximately 135 times faster with equivalent prediction performance and simpler implementations. This paper describes these algorithms in detail along with the Kalman and extended Kalman Filter predictors tested against. In addition, we describe the details of a predictor experiment and present empirical results supporting the validity of our claims that these predictors are faster, easier to implement, and perform equivalently to the Kalman and extended Kalman filtering predictors.
Vehicle crashes caused by driver distraction are of increasing concern. One approach to reduce the number of these crashes mitigates distraction by giving drivers feedback regarding their performance. For these mitigation systems to be effective, drivers must trust and accept them. The objective of this study was to evaluate real-time and post-drive mitigation systems designed to reduce driver distraction. The real-time mitigation system used visual and auditory warnings to alert the driver to distracting behavior. The post-drive mitigation system coached drivers on their performance and encouraged social conformism by comparing their performance to peers. A driving study with 36 participants between the ages of 25 and 50 years old (M¼34) was conducted using a high-fidelity driving simulator. An extended Technology Acceptance Model captured drivers' acceptance of mitigation systems using four constructs: perceived ease of use, perceived usefulness, unobtrusiveness, and behavioral intention to use. Perceived ease of use was found to be the primary determinant and perceived usefulness the secondary determinant of behavioral intention to use, while the effect of unobtrusiveness on intention to use was fully mediated by perceived ease of use and perceived usefulness. The real-time system was more obtrusive and less easy to use than the post-drive system. Although this study included a relatively narrow age range (25 to 50 years old), older drivers found both systems more useful. These results suggest that informing drivers with detailed information of their driving performance after driving is more acceptable than warning drivers with auditory and visual alerts while driving.
The aim of this study was to assess how scrolling through playlists on an MP3 player or its aftermarket controller affects driving performance and to examine how drivers adapt device use to driving demands.
Drivers use increasingly complex infotainment devices that can undermine driving performance. The goal activation hypothesis suggests that drivers might fail to compensate for these demands, particularly with long tasks and large search set sizes.
A total of 50 participants searched for songs in playlists of varying lengths using either an MP3 player or an aftermarket controller while negotiating road segments with traffic and construction in a medium-fidelity driving simulator.
Searching through long playlists (580 songs) resulted in poor driving performance and required more long glances (longer than 2 s) to the device compared with other playlist lengths. The aftermarket controller also led to more long glances compared with the MP3 player. Drivers did not adequately adapt their behavior to roadway demand, as evident in their degraded driving performance. No significant performance differences were found between short playlists, the radio-tuning task, and the no-task condition.
Selecting songs from long playlists undermined driving performance, and drivers did not sufficiently adapt their use of the device to the roadway demands, consistent with the goal activation hypothesis. The aftermarket controller degraded rather than enhanced performance.
Infotainment systems should support drivers in managing distraction. Aftermarket controllers can have the unintended effect of making devices carried into the car less compatible with driving.These results can motivate development of new interfaces as alternatives to scrolling lists.
Lane-keeping assistance systems for vehicles may be more acceptable to users if the assistance was adaptive to the driver's state. To adapt systems in this way, a method for detection of driver distraction is needed. Thus, we propose a novel technique for online detection of driver's distraction, modeling the long-range temporal context of driving and head tracking data. We show that long short-term memory (LSTM) recurrent neural networks enable a reliable subject-independent detection of inattention with an accuracy of up to 96.6%. Thereby, our LSTM framework significantly outperforms conventional approaches such as support vector machines (SVMs).
This paper introduces a driving danger-level prediction system that uses multiple sensor inputs and statistical modeling to predict the driving risk. Three types of features were collected for the research, specifically the vehicle dynamic parameter, the driver's physiological data and the driver's behavior feature. To model the temporal patterns that lead to safe/dangerous driving state, several sequential supervised learning algorithms were evaluated in the paper, including hidden Markov model, conditional random field and reinforcement learning. Experimental results showed that using reinforcement learning based method with the vehicle dynamic parameters feature outperforms the rest algorithms, and adding the other two features could further improve the prediction accuracy. Based on the result, a live driving danger-level prediction prototype system was developed. Compared to many previous researches that focused on monitoring the driver's vigilance level to infer the possibility of potential driving risk, our live system is non-intrusive to the driver, and hence it is very desirable for driving danger prevention applications. Subjective on-line user study of our prototype system gave promising results.
This study aims to develop an automatic method to detect drowsiness onset while driving. Support vector machines (SVM) represents a superior signal classification tool based on pattern recognition. The usefulness of SVM in identifying and differentiating electroencephalographic (EEG) changes that occur between alert and drowsy states was tested. Twenty human subjects underwent driving simulations with EEG monitoring. Alert EEG was marked by dominant beta activity, while drowsy EEG was marked by alpha dropouts. The duration of eye blinks corresponded well with alertness levels associated with fast and slow eye blinks. Samples of EEG data from both states were used to train the SVM program by using a distinguishing criterion of 4 frequency features across 4 principal frequency bands. The trained SVM program was tested on unclassified EEG data and subsequently checked for concordance with manual classification. The classification accuracy reached 99.3%. The SVM program was also able to predict the transition from alertness to drowsiness reliably in over 90% of data samples. This study shows that automatic analysis and detection of EEG changes is possible by SVM and SVM is a good candidate for developing pre-emptive automatic drowsiness detection systems for driving safety.
The goal of this work is the detection and classification of driver activities in an automobile using computer vision. To this end, this paper presents a novel two-step classification algorithm, namely, an unsupervised clustering algorithm for grouping the actions of a driver during a certain period of time, followed by a supervised activity classification algorithm. The main contribution of this work is the combination of the two methods to provide a computationally fast solution for deployment in real-world scenarios that is robust to illumination and segmentation issues under most conditions experienced in the automobile environment. The unsupervised clustering groups the actions of the driver based on the relative motion detected using a skin-color segmentation algorithm, while the activity classifier is a binary Bayesian eigenimage classifier. Activities are grouped as safe or unsafe and the results of the classification are shown on several subjects obtained from two distinct driving video sequences.
In this paper, we review the state-of-the-art technologies for driver inattention monitoring, which can be classified into the following two main categories: 1) distraction and 2) fatigue. Driver inattention is a major factor in most traffic accidents. Research and development has actively been carried out for decades, with the goal of precisely determining the drivers' state of mind. In this paper, we summarize these approaches by dividing them into the following five different types of measures: 1) subjective report measures; 2) driver biological measures; 3) driver physical measures; 4) driving performance measures; and 5) hybrid measures. Among these approaches, subjective report measures and driver biological measures are not suitable under real driving conditions but could serve as some rough ground-truth indicators. The hybrid measures are believed to give more reliable solutions compared with single driver physical measures or driving performance measures, because the hybrid measures minimize the number of false alarms and maintain a high recognition rate, which promote the acceptance of the system. We also discuss some nonlinear modeling techniques commonly used in the literature.
It is well established in the literature that secondary tasks adversely affect driving behavior. Previous research has focused on discovering the general trends by analyzing the average effects of secondary tasks on a population of drivers. This paper conjectures that there may also be individual effects, i.e., different effects of secondary tasks on individual drivers, which may be obscured within the average behavior of the population, and proposes a model-based approach to analyze them. Specifically, a radial-basis neural-network-based modeling framework is developed to characterize the normal driving behavior of a driver when driving without secondary tasks. The model is then used in a scenario of driving with a secondary task to predict the hypothetical actions of the driver, had there been no secondary tasks. The difference between the predicted normal behavior and the actual distracted behavior gives individual insight into how the secondary tasks affect the driver. It is shown that this framework can help uncover the different effects of secondary tasks on each driver, and when used together with support vector machines, it can help systematically classify normal and distracted driving conditions for each driver.
Emotional human–computer interactions are attracting increasing interest with the improvement in the available technology. Through presenting affective stimuli and empathic communication, computer agents are able to adjust to users' emotional states. As a result, users may produce better task performance. Existing studies have mainly focused on the effect of only a few basic emotions, such as happiness and frustration, on human performance. Furthermore, most research explored this issue from the psychological perspective. This paper presents an emotion and performance relation model in the context of vehicle driving. This general emotion–performance model is constructed on an arousal–valence plane and is not limited to basic emotions. Fifteen paid participants took part in two driving simulation experiments that induced 115 pairs of emotion–performance sample. These samples revealed the following: (1) driving performance has a downward U-shaped relationship with both intensities of arousal and valence. It deteriorates at extreme arousal and valence. (2) Optimal driving performance, corresponding to the appropriate emotional state, matches the “sweet spot” phenomenon of the engagement psychology. (3) Arousal and valence are not perfectly independent across the entire 2-D emotion plane. Extreme valence is likely to stimulate a high level of arousal, which, in turn, deteriorates task performance. The emotion–performance relation model proposed in the paper is useful in designing emotion-aware intelligent systems to predict and prevent task performance degradation at an early stage and throughout the human–computer interactions.
Crash causation research has identified inattention as a major source of driver error leading to crashes. The series of experiments presented herein investigate the characteristics of an in-vehicle information system (IVIS) task that could hinder driving performance due to uncertainty buildup and cognitive capture. Three on-road studies were performed that used instrumented passenger and tractor-trailer vehicles to obtain real-world driving performance data. Participants included young, middle-aged, and older passenger vehicle drivers and middle-aged and older commercial vehicle operators. While driving, they were presented with IVIS tasks with various information densities, decision-making elements, presentation formats, and presentation modalities (visual or auditory). The experiments showed that, for both presentation modalities, the presence of multiple decision-making elements in a task had a substantial negative impact on driving performance of both automobile drivers and truck drivers when compared to conventional tasks or tasks with only one decision-making element. The results from these experiments can be used to improve IVIS designs, allowing for potential IVIS task phenomena such as uncertainty buildup and cognitive capture to be avoided.
We present a non-intrusive prototype computer vision system for real-time monitoring driver's vigilance. It is based on a hardware system, for real time acquisition of driver's images using an active IR illuminator, and their software implementation for monitoring some visual behaviours that characterize a driver's level of vigilance. These are the eyelid movements and the pose face. The system has been tested with different sequences recorded on night and day driving conditions in a motorway and with different users. We show some experimental results and some conclusions about the performance of the system.
We present an experiment comparing double exponential smoothing and Kalman filter-based predictive tracking algorithms with derivative free measurement models. Our results show that the double exponential smoothers run approximately 135 times faster with equivalent prediction performance. The paper briefly describes the algorithms used in the experiment and discusses the results.