Monitoring driving status has great potential in helping us decline the occurrence probability of traffic accidents and the aim of this research is to develop a novel system for driving stress detection based on multimodal feature analysis and kernel-based classifiers. Physiological signals such as electrocardiogram, galvanic skin response and respiration were record from fourteen drives executed in a prescribed route at real drive environments. Features were widely extracted from time, spectral and wavelet multi-domains. In order to search for the optimal feature sets, Sparse Bayesian Learning (SBL) and Principal Component Analysis (PCA) were combined and adopted. Kernel-based classifiers were employed to improve the accuracy of stress detection task. Analysis I used features from 10 s intervals of data which were recorded during well-defined rest, highway and city driving conditions to discriminate three levels of diving stress achieving an averaging accuracy over 99% at per-drive level and 89% in cross-drive validation. Analysis II made continuous stress evaluation throughout a complete driving test attaining a high coincidence with the true road situation especially at the switching interval of traffic conditions. Experimental results reveal that different levels of driving stress can be characterized by specific set of physiological measures. These physiological measures could be applied to in-vehicle intelligent systems in various approaches to help the drivers better manage their negative driving status. Our design scheme for driving stress detection could also facilitate the development of similar in-vehicle expert systems, such as driver's emotion management, driver's sleeping onset monitoring, and human-computer interaction (HCI).

Detecting driving stress in physiological signals based on multimodal feature analysis and kernel classifiers

Many traffic injuries and deaths are caused by the drowsiness of drivers during driving. Existing drowsiness detection schemes are not accurate due to various reasons. To resolve this problem, an accurate driver drowsiness classifier (DDC) has been developed using an electrocardiogram genetic algorithm-based support vector machine (ECG GA-SVM). In existing studies, a cross correlation kernel and a convolution kernel have both been applied for performing the classification. The DDC is designed by a Mercer kernel ${{\rm{K}}_{{\mathrm{DDC}}}}$ formed by commuting the cross correlation kernel ${{\rm{K}}_{{\mathrm{xcorr}},\mathrm{ij}}}$ and the convolution kernel ${{\rm{K}}_{{\mathrm{conv}},\mathrm{ij}}}$ . ${{\rm{K}}_{{\mathrm{xcorr}},\mathrm{ij}}}$ , and captures the symmetric information among ECG signals from different classes, while ${{\rm{K}}_{{\mathrm{conv}},\mathrm{ij}}}$ captures the antisymmetric information among ECG signals from the same class. The final ${{\rm{K}}_{{\mathrm{DDC}}}}$ (a precomputed kernel) is obtained by a genetic mutation using a multi-objective genetic algorithm. This renders an optimal ${{\rm{K}}_{{\mathrm{DDC}}}}$ that confidently serves as the full descriptor of the drowsiness. The performance of ${{\rm{K}}_{{\mathrm{DDC}}}}$ is compared with the most prevailing kernels. The obtained DDC yields an overall accuracy of 97.01%, sensitivity of 97.16%, and specificity of 96.86%. The analysis reveals that the accuracy of ${{\rm{K}}_{{\mathrm{DDC}}}}$ is better than those of both ${{\rm{K}}_{{\mathrm{xcorr}},\mathrm{ij}}}$ and ${{\rm{K}}_{{\mathrm{conv}},\mathrm{ij}}}$ by more than 11%, and typical kernels including linear, quadratic, third order polynomial, and Gaussian radial basis function by 17–63%, respectively. Comparing with related works using the image-based method and the biometric signal-based method, K DDC improves the accuracy by 48.4–87.2%. Testing results showed that KDDC has a less than 1% deviation from simulated results. Also, the average delay of DDC was bounded by 0.55 ms. This renders the real time implementation. Thus, the developed ECG GA-SVM provides an accurate and instantaneous warning to the drivers before they fall into sleep. As a result this ensures the public transport safety.

An Accurate ECG-Based Transportation Safety Drowsiness Detection Scheme

A survey on context-aware vehicular network applications

In this paper we develop a model capable of classifying drivers from their driving behaviors sensed by only low level sensors. The sensing platform consists of data available from the diagnostic outlet (OBD) of the car and smartphone sensors. We develop a window based support vector machine model to classify drivers. We test our model with two datasets collected under both controlled and naturalistic conditions. Furthermore, we evaluate the model using each sensor source (car and phone) independently and combining both the sensors. The average classification accuracies attained with data collected from three different cars shared between couples in a naturalistic environment were 75.83%, 85.83% and 86.67% using only phone sensors, only cars sensors and combined car and phone sensors respectively.

Driver Classification Based on Driving Behaviors

The problem addressed in this paper is the detection of drowsiness based on signals acquired using non-invasive and wearable sensors. A combination of signals from electroencephalography (EEG) electrodes, accelerometer, and gyroscope is acquired through a commercially available off-the-shelf headband to achieve this goal. The aim is to substitute the use of intrusive sensors and cameras by extracting both physiological and behavioral features for drowsiness detection using a lightweight wearable headband. A dataset is acquired from 50 volunteers using the MUSE headband for both fresh and drowsy behavior in the resting state, where subjects stare at a slow moving object on a screen and no external stimuli is used. The drowsy state is induced by sleep deprivation. The data are labeled subjectively using the Karolinska sleepiness scale. Complementary features extracted from spectral signals, head movement, and blink analysis are combined to form a feature vector for classifiers to improve the robustness and detection accuracy of the system. The implemented system is validated on the data collected in fresh and drowsy states using a leave-one-out cross-subject and 25% hold-out level testing. A detection accuracy of 92% is achieved using selected features and a linear support vector machine classifier. The results show that the accuracy of drowsiness detection is improved by using a hybrid feature vector acquired from data using multiple sensors.

Wei Zhang

Papers

Driver drowsiness recognition based on computer vision technology

Digital speckle technique applied to flow visualization