Aims of this study are to assess the effect of warning on behavioral performance and brain dynamics and to determine the sustainable duration of warning in terms of behavioral responses and neural activities. Results showed that the warning sounds can accelerated the responses to the car drifting and suppress the tonic and phasic power increases at the alpha and theta bands in the occipital area. Furthermore, the effects of the warning sounds on the behavior and brain activities could sustain to 12.6±2.7 and 10 sec respectively.
4.1. Effects of drowsiness
Behavioral results showed that the sorted RT curves of drowsy trials, including the warning and without warning trials were all significantly departed from the sorted RT curves of alert current. This finding was consistent with previous studies suggested that the subjects’ behavioral performance was decreased along with changes of cognitive states from alert to drowsy [13].
The decremented behavioral performances could be assessed in terms of response time or driving trajectory [8, 9].
Our results also showed that the neural activities were also altered with the cognitive states. Specifically, both the tonic and phasic changes of the occipital area showed that the alpha and theta band power were increased in the drowsy relative alert trials. This finding was consistent with previous reports and our previous studies. For example, the tonic EEG power was higher on average in the drowsiness than in the alertness. In addition, the
increases of power in the low-theta frequencies near 4 Hz were highly correlated with the drowsiness [59, 65]. Huang et al. also reported the baseline spectra had tonic broad band power increases at the occipital area during periods of relatively poor driving performance and the prominent spectral changes were observed in alpha and theta bands [60]. Taken together, our results demonstrated that our experimental paradigm could successfully induced the drowsiness under driving condition in terms of behavioral performances and changes of brain rhythm at the occipital region. In addition, the warning sounds randomly delivered to the subjects did not affect the drowsiness induction during the experiment.
Except the occipital region, the power of the EEG spectra for central, left and right somatomotor and the parietal regions were increased around frequencies at 3-12 Hz (alpha and theta band) and 20-25 Hz (beta band).
Results showed that the power increases at alpha, beta and theta bands associated with the drowsiness could be involved in large brain areas. Such phenomenon may due to the drowsy related brain dynamics was probably modulated by the independent modulators. The concept of the independent modulator hypothesized that the independent modulator could modulate the brain oscillations across several distinct cortical areas [66]. Effects of warning on altering the brain oscillations did not observed in other brain area. Such results suggested that those changes of brain dynamics were not due to the neural responses to the sounds.
4.2. Effects of warnings
4.2.1. Behavior performance
Results showed that warning sounds improved the subject’s behavior performance by accelerating their RT. Specifically, the RTs of the current trials were significant shorten about 1.03 nu in the warning relative to without warning trials. The effects of warning on enhancing the behavioral performance also sustained to the next trails. Specifically, the warning could accelerate the RT about 0.37 nu faster than the RT of trials without warning.
Lin et al. [43] showed that the mean RT of sessions with warning was significantly faster than that of sessions without warning. The mean RT was reduced by approximately 1.15 seconds. Several studies also reported that the warning sounds could help drivers to react promptly [67, 68] and reduce the probability of collision [68, 69].
Comparing with the RTs of alertness, the intervals between the onset of warning signals and the subject’s responses were still longer than the RTs of alert trials. The results implied that the warning signals could enhanced the behavioral performance but the subject’s conscious could be not as clear as the alert. The reasons may relate to the characteristics of warnings sound is not ideal or the single stimulus may not be enough to awake the subject. Study suggested that the characteristic of most powerful sounds to awake the drowsy driver is auditory icon, such as the horn or tire skid [40, 41]. In addition, auditory neuron has know easily adapt to the pure tone or pure tone burst [70].
In the auditory cortices, the majority of neurons are only responses to the complex sounds and only few neurons can respond to the pure tone or pure
tone burst [71]. Previous studies also suggested that warning signals delivered by the single modality may not sufficient to total awake the subjects and they suggested that warning signals delivered through the multimodalities, such as the combined the warning sounds and vibrations, could be a better methods to keep drivers’ alertness [72].
4.2.2. Brain dynamics altered by the warning sounds
The significantly inhibited the tonic increases of the mean baseline power were observed in the warning trials comparing with those trails without warning.
The suppressed brain oscillations were mainly found at the theta and alpha bands, which are widely used as drowsiness related features [27]. The phasic decreases of the theta-band and alpha-band power were also observe in the ERSPs around the onset of the warning sounds and such the decreased of theta and alpha band power could sustain at least for around 10 sec. The findings suggested that that warning sounds would help drivers to reduce the drowsiness reflected on both behavioral performance and brain oscillations.
This is the first study to show that the warning feedback could partially change the driver’s cognitive states. Consistent with the behavioral results, the neurophysiologic data also showed that the warning sounds couldn’t totally remove the driver’s drowsiness. In addition to the non-ideal characteristics and presentation modality of warning signals, such results may also relate to the effectiveness of warning feedback may decrease along with the increases of drowsiness. Comparing effects of the warning on altering the brain activities among three different RT groups, results suggested that the warning sounds could be effective when the normalized RTs was less than 4 nu, which indexed
block all the sensory inputs during the sleep [73]. It is still unclear that the thalamus gate would be digitally blocked or decreased the sensory inputs analogically from drowsy to sleep. According to our preliminary results, we speculated that the sensory inputs probably may be attenuated along with the degradation of the alertness. The control mechanisms of thalamus gate in details need to further assess in the future.
4.3. Duration of the warning effects
The alpha- and theta bands showed different intervals of phasic power decreases. The duration of phasic alpha band decreases was around 10 sec, while the theta-band power decrease was around 35 sec. Such difference may due to the alpha band power have nonlinear and relative larger fluctuations in the theta band power during the transition from alert to drowsy. Studies showed the alpha activities increased and then started to decrease during wake-sleep transition [74-76]. Chuang et al. also reported that during mild drowsy period, the fluctuation of alpha activations was larger than theta and beta fluctuation [66]. Such alpha fluctuations could result from event-related desynchronization (ERD) and synchronization (ERS) of alpha activities [48, 60, 77] during the responses to the car deviation by manipulating the steering wheel.
The effects of warning sound can sustain for a short duration around 10 sec. This implies that the warning sounds could only transiently change the driver’s drowsiness. For the safety concerned, it is necessary to either combine with other feedback methods or to include the automatically driving system in the future to help the drowsy drivers to avoid the car crashes.