Studyof Driver’s Cerebral Functional Network Based On Near-infrared Spectroscopy

With the development of new technology, more and more information interactive systems are added into vehicle cab. As secondary missions, these systems cost driver’s brain resources and raise the level of driver’s mental workload, while making it quite convenient for drivers to acquire information and operate the vehicles. Nowadays most of the assessment methods of cab human-machine interaction interface just test the user’s experience and convenience but include no test about the influence of interaction interface on driver’s mental workload. Therefore, this thesis aims to develop a model of cerebral functional network which can quantitatively evaluate driver’s mental workload level. These indices calculated from the model can be used to check whether the design of vehicle cab is rational, as well as provide improvement suggestions.This research uses multi-channel near-infrared spectroscopy (NIRS) method to record concentration changes of oxy-hemoglobin in multiple cerebral areas, which reflect the action of cerebral cortices. Then, by combining wavelet-based coherence method and graph theory, this thesis develops a new cerebral functional network mode. The channel in NIRS and coherence between signals are defined as vertex and edge in graph theory, respectively. After this definition, adjacent matrix and weighted matrix of complex network could be calculated and clustering coefficient, local efficiency and global efficiency, which are 3 typical indices used to assess network character, could be obtained. To distinguish different fluctuation components caused by different physiological activities, this study divides the whole frequency range of blood oxygenation signal into 6 frequency bands and analyze the cerebral functional network character in each band.In order to validate whether the new model can reflect true features of cerebral functional network, this research conducts a contrast experiment between healthy elderly subjects and elderly subjects with cerebral infarction (CI) during resting state. A 10-channel NIRS system is used to record the oxygenation oscillations of each subject in prefrontal and motor cortex. Moving average and Butterworth filter is used to preprocess each signal and then coherence between every two signals is calculated to finally obtain the adjacent matrix and weighted matrix of each subject. Results show that all 3 indices of CI group are significantly lower in III, IV and V frequency intervals than healthy group. This result demonstrates that the new model could reflect characters of the cerebral functional network more comprehensively.Finally, this thesis applies the new model to detect driver’s mental workload level by conducting a driving task. A 6-channel NIRS system is used to record the oxygenation oscillations of each subject in prefrontal and motor cortex. Results of the experiment show that all 3 indices of the new model during driving task are significantly lower in III, IV or IV, V intervals than resting state. The results indicate that driving may cause higher mental workload and result in lower cerebral functional network indices.In summary, the new model in this thesis could reflect characteristics of the cerebral functional network and could quantitatively evaluate the mental workload level of driver. Applying this model to design of vehicle cab human-machine interaction interface could detect the influence of these interactive systems on drivers, and provide suggestions to improve cab design, which is important for driving safety.