Toshihito Ikenishi

Parallel Factor Analysis of Driver's EEG in Lane Change Maneuver for Cooperative Driver Assistance Systems

Abstract Vehicle technology of the better interaction between human and machine has been called human-electronics in Japan. It is necessary to obtain better relationship between human and vehicle. A drivers information, which can be obtained from steering operation, pedal operation, camera images and physiological information, particularly is important to find a method to determine a drivers operational intention. In terms of using this physiological information, an area of focus is using brain activity. Recently, some former researches haves been reported about the investigation of the brain activity of the driver. The traditional decomposition of the electroencephalograph (EEG) has been based on the two-dimension components. In the multiple electrode analysis of EEG, the frequency-spatial domain and the time-spatial domain have been used. However, these conventional methods can only use two-dimensional data. As a multi-channel EEG analysis using multi dimensional data, parallel factor analysis (PARAFAC) method is based on the report. In this paper, we described that the drivers EEG during lane change was decomposed by PARAFAC and we investigated the factor of recognize and judgment from that decomposition result. Consequently, Common to the all subjects has 2 factors which were in the 5-8 Hz and 8-13 Hz. Those factors considered that they were changed by the drivers mental state, during recognition for alpha wave and during judgment for theta wave.

Estimation of Driver's Longitudinal Intention for the Preceding Car by Brain Current Distribution Estimation Method

The interaction between human and vehicle has been extensively researched to reduce the traffic accident in Japan. In the research of driver assistance system, human-friendly driver assistance systems have been researched using the information of driver and vehicle. This system requires to achieve a better relationship between human and vehicle. In addition, it is important to find a method to detect drivers operational intention. Therefore, we have focused on the brain activities in the biological information. In our previous research, we investigated that the drivers EEG at the preceding car avoidance maneuver was decomposed by parallel factor analysis (PARAFAC), and we investigated the drivers EEG of during longitudinal operation. Consequently, all subjects have two common factors of the frequency component which exist in the 5-10 Hz and 8-13 Hz bandwidth. Those factors were changed by the drivers mental state during visual recognition and judgment. In this paper, we estimated the drivers longitudinal intention from a drivers EEG using source current distribution estimation with Hierarchical Bayesian method and the sparse logistic regression. From the estimation results, the estimation accuracy of drivers intention was higher than about 60 % accuracy of all operation except the gas pedal operation of ones subject.

Analysis of longitudinal driving behaviors during car following situation by the driver's EEG using PARAFAC

Abstract Vehicle technology of the interaction between human and the machine has been called human-electronics in Japan. It is necessary to achieve a better relationship between human and vehicle. A drivers information, which can be obtained from steering operation, pedal operation, camera images and physiological information, particularly is crucial to find a method to determine a drivers operational intention. Recently, some former researches have been reported about the investigation of the brain activity of the driver. The time frequency analysis such as FFT has been major method in the traditional decomposition of the electroencephalogram (EEG). However, these conventional methods can only use two-dimensional data. In this paper, we described that the drivers EEG during car following was decomposed by parallel factor analysis (PARAFAC), and we investigated the feature factor of longitudinal behavior for recognize and judgment from that decomposition result. PARAFAC analysis has known as a multi-channel EEG analysis of multi-dimensional data. Consequently, Common to all subjects has two factors of the frequency component which were in the 5-10 Hz and 8-13 Hz. Those factors were changed by the drivers mental state during visual recognition and judgment. In addition, we estimated the feature factor from a new EEG data set using inverse solution of PARAFAC. From estimation results, the driver recognized preferentially shape and color than distance and movement information in the car following situation.

Feature analysis of acceleration and deceleration behavior during hurry driving using driving behavior model based on hierarchical Bayesian approach

Abstract Existing driver assistance systems are based on averaged characteristics of drivers so the systems may cause a sense of discomfort to some drivers and the effectiveness on accident prevention degrades due to low system acceptance. To deal with this issue, an individual adaptive hurry driving detection system is proposed in this research. We proposed the detection method of longitudinal hurry driving using hierarchical Bayesian model. Urban driving data are collected by a continuous-logging drive recorder (DR). The system using Hierarchical Bayesian method has extracted feature of hurry driving behavior appropriately. These features were different of individual drivers. The probability of hurry driving state is detected by using this hierarchical Bayesian model.