Effects of Driving Events

Real-life driving usually involves a lot of motor activities, including steering, braking, stepping on the gas paddle, and shifting gears, in response to events or situations on the road. These events and responses generally induce transient (phasic) perturbation in the EEG power spectra, which are referred to as event-related spectra perturbations (ERSPs) [17][26]; however, it is not known whether these phasic activities may affect the trends in EEG power spectra and the validity of drowsiness detection. In Chapter IV, tonic EEG power spectra were only computed from a 1-sec window prior to deviation onset in each epoch, which was less contaminated by phasic activities; in this section, EEG power spectra from the whole 8-sec epoch, which contained the effects of different events (deviation onset, response onset, and response offset), were estimated using the same methods described in Section 3.6 and Section 3.7. The discussions in this section suggest that phasic activities did influence the trends in some frequency bands of some IC clusters. The goal for developing a reliable drowsiness detection system is to

iden-Only motion datasets are discussed since the experiment conditions are closer to real-life driving than their motionless counterparts.

5.3.1 The Frontal Cluster

Figure 40 shows the comparison between “tonic” (only the pre-deviation period in each epoch) and “mixed” (both pre- and post-deviation period in each epoch) power spectra of the frontal cluster. The power images (Figure 40 A) show that in-crease in “mixed” power spectra is less salient than that in the “tonic” ones; however, significant theta band power increase occurs at lower RTs in the “mixed” power spectra. The trends in Figure 40 B show the “mixed” power increase in the theta band is lower than that of “tonic” power increase at longer RTs, but the monotonic rising trend is not affected by phasic activities in the “mixed” power spectra, sug-gesting that the theta band power in the frontal cluster could be a stable index for drowsiness detection in real-life driving.

Figure 40: Comparison of the trends between “tonic” and “mixed” power spectra of the frontal cluster.

A: moving averaged power images (left panel: “tonic” power spectra, right panel: “mixed” power spectra; regions inside contour: p < 10^-10 by two-tailed t-test, corrected). B: trends of power in-crease in four frequency bands extracted from A; light-blue traces: trend in “tonic” power, light-green traces: trend in “mixed” power).

5.3.2 The Central and Parietal Clusters

The comparison between the trends in “tonic” and “mixed” power spectra in the central cluster is shown in Figure 41. The “mixed” power image shows significant increase in both theta and beta bands at long RTs (> 2.5 sec), and the overall

“mixed” power in the theta and alpha bands is lower than that in the “tonic” power spectra. The rising trend of theta band power in the “mixed” power spectra suggest that theta band power is less affected by “phasic” activities and may be suitable for drowsiness detection in real-life driving.

Figure 41: Comparison of the trends between “tonic” and “mixed” power spectra of the central cluster.

Other conventions follow Figure 40.

Figure 42 shows the comparison between “tonic” and “mixed” power spectra in the parietal cluster. The power images show almost no changes in “mixed” power spectra except theta band power when RTs are longer than ~2.22 sec. The trends of

“mixed” power only show small increase in the theta band and remain unchanged in other frequency bands. These results suggest that “mixed” power spectra of the pa-rietal cluster are largely affected by “phasic” activities, and may not be suitable for drowsiness detection.

Figure 42: Comparison of the trends between “tonic” and “mixed” power spectra of the Parietal clus-ter. Other conventions follow Figure 40.

5.3.3 The Somatomotor Clusters

Figure 43 and Figure 44 show the comparisons of trends between “tonic” and

“mixed” power spectra in the left and right somatomotor clusters. The power images and trends of the left somatomotor cluster show the changes in “mixed” power spectra are generally smaller than those in the “tonic” power spectra; moreover, alpha and beta band power is further suppressed with increasing RTs. The power images and trends of the right somatomotor cluster do not show consistent differ-ence between “tonic” and “mixed” power spectra in four frequency bands. These results suggest that the trends of power spectra of the somatomotor clusters are affected by motor responses and are not suitable for drowsiness detection.

Figure 43: Comparison of the trends between “tonic” and “mixed” power spectra of the left somato-motor cluster. Other conventions follow Figure 40.

Figure 44: Comparison of the trends between “tonic” and “mixed” power spectra of the right soma-tomotor cluster. Other conventions follow Figure 40.

5.3.4 The Occipital Clusters

Figure 45-Figure 47 show the comparisons of trends between “tonic” and

“mixed” power spectra in three occipital clusters. The power images and trends show that the overall power in the “mixed” power spectra is generally lower than that of the “tonic” power spectra in all three occipital clusters. In particular, alpha band power is greatly reduced in the “mixed” power spectra, which could be explained by the subject’s attention and responses following deviation onset. The trends of theta band power in the “mixed” power spectra remain virtually the same as those in the

“tonic” power spectra; in addition, significant power increases occur at longer RTs in the “mixed” power spectra. These results suggest that theta band power in the

oc-Figure 45: Comparison of the trends between “tonic” and “mixed” power spectra of the occipital mid-line cluster. Other conventions follow Figure 40.

Figure 46: Comparison of the trends between “tonic” and “mixed” power spectra of the bilateral oc-cipital cluster. Other conventions follow Figure 40.

Figure 47: Comparison of the trends between “tonic” and “mixed” power spectra of the tangential occipital cluster. Other conventions follow Figure 40.