The purpose of this study is to investigate continuous EEG fluctuations from alertness to drowsiness in a realistic VR based driving environment. Firstly, we applied ICA to the EEG collected from each individual separately. Then we clustered the EEG sources from all the volunteer participants based on their scalp map gradients. Seven component clusters were identified: BLO, OM, FCM, CM, CPM, LCP and RCP clusters. Secondly, time-frequency analysis is used to assess consistent EEG correlates of the cognitive-state changes across subjects. The results show that alpha band power increase during the transition from alertness to very-slight and slight drowsiness, but remain constant or slightly decrease during extreme drowsiness period for all the clusters. On the other hand, the theta band power of each component cluster increased monotonically during the transition from slight to extreme drowsiness. The experimental results show that previous studies might just investigate parts of transition from alertness to drowsiness.
5.1 The EEG Fluctuations from Alertness to Drowsiness
The EEG fluctuations from alertness to drowsiness during this experiment were comparable to the results that reported in previous studies [46]. In traditional sleep EEG studies, the alpha-power decrease and theta-power increase were the EEG characteristics of sleep stage 1 (also called “Drowsiness”) and microsleep [42-43, 47-48]. It is similar to the results in our extreme drowsiness periods. This study focuses on the cognitive-state transition during wakefulness and finds the theta power not only increases from wakefulness to sleep stage 1 but also from alertness to drowsiness.
The alpha rhythm is the first defined EEG rhythm (Berge, 1929). EEG synchronization within the alpha band is an electrophysiological correlate of cortical idling [49-50]. The areas that are not processing sensory information or motor output can be considered to be in an
idling state. Therefore, drowsiness could an idling state of the brain.
The trends of alpha- and theta-power changes from alertness to drowsiness were similar between different brain regions. Additionally, Table III also shows that high percentage of subjects with high correlation between powers of the different components. Hence, the drowsy related alpha and theta rhythm in these components maybe modulated by the same nucleus [51-55]
5.2 Lane-Keeping Driving Task related Cerebral Cortex
According the Brodmann’s map, the BLO and OM clusters were located in visual cortices (Area 17, 18, a.k.a. V1, V2). The V1 cortex is the simplest, earliest cortical visual area. It is highly specialized for processing information about static and moving objects and was excellent in pattern recognition [56-57]. It seems physiologically feasible that V1 which includes very large attentional modulation [58-59] involves in this task. V2 was the second major area in the visual cortex. It received strong feedforward connections from V1 and sended strong connections to V3, V4, and V5. Functionally, V2 had many properties in common with V1 and recent research had shown that V2 cells exhibit a small amount of attentional modulation [59]. Therefore, the inclusion of V2 in this lane-keeping driving task seems also plausible.
The CPM component cluster was located near areas 7 and 19 (V3). V3 was a term used to refer to the regions of cortex located immediately in front of V2. V3 can be divided into two subareas, dorsal V3 and ventral V3. Dorsal V3 was normally considered to be a part of the dorsal stream. Recent work with fMRI had suggested that area V3 may play a role in the processing the information of global motion [60]. The area 7 was a somatosensory association cortex that involves in locating objects in space. It served as a point of convergence between vision and proprioception to determine where objects are in relation to parts of the body [61-62]. In this study, the experimental setup was based on 360o VR technology. When the car
drifted in this VR environment, the subject received the motion and spatial information during the epxeriments, which might explain the involvement of Aera 7.
The LCP, RCP and CM clusters located near Brodmann areas 1, 2, 3, 4 and 6. The areas 1,2 and 3 are also called primary somatosensory cortex which consists of the various sensory receptors that trigger the experiences labelled as touch or pressure, temperature (warm or cold), pain (including itch and tickle), and the sensations of muscle movement and joint position including posture, movement, and facial expression (collectively also called proprioception) [42]. Areas 4 and 6 are primary motor cortex and pre-motor cortex which plan and execute movements [63-64]. In this study, the subject needs to respond to lane deviation by steering the wheel. Therefore, the muscle movement and joint position including posture were sensed by somatosensory cortex. Whereas the action of steering wheel were planned and executed by pre-motor cortex and primary motor cortex.
The FCM cluster located near Brodmann areas 9 and 46 that play a role in sustaining attention and working memory [65]. In our study, the subject needs to keep attention on the lane-keeping driving task. Therefore, the attentional network unavoidably involved in the task.
5.3 The Fluctuations of EEG Alpha and Theta Power for Detecting Driver’s Drowsiness
In previous studies that suggested the use of EEG signals is potentially the best for detecting vigilance while driving. [66-68] In the present study, we compared the EEG between different component clusters diversity of EEG power changes with respect to the transition from alertness to drowsiness and found that alpha power of BLO and OM component were most stable and desirable EEG feature for very-slight drowsiness detection.
Additionally, the alpha and theta power of BLO and OM component were most stable and desirable EEG feature for slight drowsiness detection. Lastly, the theta power of BLO and
OM component were most stable and desirable EEG feature for extreme drowsiness detection.
As the characteristic of drowsiness related EEG activity described above, if a person is very-slight drowsiness, the alpha wave will tend to be superior in EEG activity, and its power will increase time after time in occipital lobe, remarkably. After that, if the person tends to fall slight drowsiness, the power of alpha and theta will increase time after time in occipital lobe, remarkably. After that, if the person tends to fall extreme drowsiness, the power of alpha will decrease while the theta will still increase time after time in occipital lobe. With these results, we can quantify the driver’s consciousness level based on their EEG activity in the frequency domain of occipital lobe. Additionally, we also can develop an alarm system for motor vehicle crash prevention.