In this thesis, passengers' brain signals, including electroencephalography (EEG) and near-infrared spectroscopy (fNIRS), were analyzed to extract road information to potentially prevent car accidents and provide public trust in high-level autonomous vehicles. For the EEG part, event-related potential (ERP) and machine learning techniques were used to analyze and classify the signals of two road events. Results show that the responses are 454 ± 234 ms before the reaction, and the average recognition accuracy of the regularized linear discriminant analysis (RLDA) classifier reached 95.81%. For the fNIRS part, a quantification method, which is based on cerebral oxygen exchange in the prefrontal cortex of passengers and a risk field is introduced. We also verified our findings in a real-car automatic emergency braking and cut-in experiment. Overall, the results illustrate that EEG-based human-centric assistant driving systems have the potential of being deployed in autonomous vehicles to enhance the safety of passengers.