A Context-Aware EEG Headset System for Early Detection of Driver Drowsiness

Abstract

Driver drowsiness is a major cause of mortality in traffic accidents worldwide. Electroencephalographic (EEG) signal, which reflects the brain activities, is more directly related to drowsiness. Thus, many Brain-Machine-Interface (BMI) systems have been proposed to detect driver drowsiness. However, detecting driver drowsiness at its early stage poses a major practical hurdle when using existing BMI systems. This study proposes a context-aware BMI system aimed to detect driver drowsiness at its early stage by enriching the EEG data with the intensity of head-movements. The proposed system is carefully designed for low-power consumption with on-chip feature extraction and low energy Bluetooth connection. Also, the proposed system is implemented using JAVA programming language as a mobile application for on-line analysis. In total, 266 datasets obtained from six subjects who participated in a one-hour monotonous driving simulation experiment were used to evaluate this system. According to a video-based reference, the proposed system obtained an overall detection accuracy of 82.71% for classifying alert and slightly drowsy events by using EEG data alone and 96.24% by using the hybrid data of head-movement and EEG. These results indicate that the combination of EEG data and head-movement contextual information constitutes a robust solution for the early detection of driver drowsiness.

Keywords: EEG; driver drowsiness detection; gyroscope; mobile application; slightly drowsy events.

Figure 1

(a) Block diagram of the proposed system; (b) The top view of the designed headset; (c) The full view of the headset; (d) The fabricating procedure using 3D printer; (e) The fabricated headset prototype.

Figure 2

Structure of EEG bio-potentials conditioning circuit and a pair of commercial dry electrodes used. The locations marked by red dotted circles are the locations EEG electrodes attached in this study.

Figure 3

The classification model building chain containing three different tasks: data collection, classification validation and classification optimization.

Figure 4

The procedure of LOO cross-validation and optimization, where S_i means the i-th subject and Accuracy_i means the classification accuracy for i-th Round.

Figure 5

The working flowchart of smartphone which consists of two Activities (User interfaces) and Services (The functions running in background).

Figure 6

Comparison of EEG signals (line chart) and band power (pie chart) between wet electrodes (the top) and dry electrodes (the bottom). For the line chart, X-axis indicates the time (second). Y-axis indicates the amplitude of the digitalized EEG samples which are already filtered by the digital band-pass filter (4–30 Hz) in SPU. For the pie chart, the X-axis indicates the frequency ranged from 0–64 Hz (half of the sampling rate 128 Hz). Y-axis indicates the magnitude of FFT power.

Figure 7

Box-Whiskers plots of (a) EEG and (b) gyroscope features. The boxes have three lines to present the values for first quartile (the bottom), median, and third quartile (the top) for column data. The length between the first quartile (Q1) and the third quartile (Q3) is called interquartile range (IQR). Two addition lines at both ends of the whisker indicate the Q1 − 1.5 × IQR and Q3 + 1.5 × IQR value of a column data. Any data not included between the whiskers is plotted as outliers represented by “o” for mild outliers and “*” for extreme outliers. The number next to the outlier is the number of the data in that column, called case number; (c) ROC curve showing sensitivity (possibility of true drowsy event) and 1-specificity (possibility of false drowsy event) for extracted EEG and gyroscope features.

Figure 8

The typical slightly drowsy symptom, yawning, captured by video as well as EEG and gyroscope from a representative subject. The blue line charts represent EEG raw signals, X-axis, Y-axis and Z-axis signal of the gyroscope, respectively. The two bars on the right side of the line charts represent the EEG RBP features (the top) and the gyroscope MP features (the bottom).

Figure 9

Screenshot of the Android smartphone application that shows the EEG and gyroscope features, the estimation of the driving status, and the raw data of EEG and 3-axis gyroscope.