To appear in the
INTERNATIONAL JOURNAL OF PSYCHOPHYSIOLOGY
Changes in EEG alpha power during simulated driving: a demonstration.z Mark A Schier* Centre for Biomedical Instrumentation, School of Biophysical Sciences and Electrical Engineering, Swinburne University of Technology, Melbourne, Australia
Received 17 November 1998: received in revised form 17 December 1999; accepted 21 December 1999
Abstract The aim was to assess the suitability of EEG-based techniques to recording activity during a driving simulation task. To achieve this, an inexpensive driving simulator (comprising steering wheel, pedals and gear shift) were made to function with a personal computer running “Need for Speed” simulation software. Simulators of this type are both inexpensive and relatively realistic. The EEG was recorded from four sites on the scalp (P3, P4, F3, F4) for two laps during the driving task, and during a replay task. The driving task involved participants driving a vehicle on a simulated undulating, sealed surface circuit, without any other vehicles present. Two men were participants in this experiment. Power spectra were computed and integrated to produce values of relative alpha activity for each channel and recording epoch, a time-series of alpha activity during each recorded segment. Overall values for alpha activity indicated an increase for replay compared to driving, and also driving on lap 5 compared to driving lap 2. The EEG changes are consistent with the notion of overall reduction of attention during the later laps and the replay task and indicate the potential of such measures for complex motor behaviour. Keywords: EEG, alpha activity,
Parts of this work were presented at the 6th Australasian Conference on Psychophysiology, Hobart, Australia, 1996, and the 9th World Congress of the International Organization for Psychophysiology, Taormina, Sicily 1998. * School of Biophysical Sciences and Electrical Engineering, Swinburne University of Technology PO Box 218 Hawthorn VIC 3122, Australia Tel +61-3-9214-8713: fax +61-3-9819-0856 E-mail address:
[email protected]
z
© Copyright of the published article is held by Elsevier Science, Rights and Permissions Department, PO Box 800 Oxford, OX5 1DX UK. This preprint may be reproduced in an intact form. The final published article appears in: International Journal of Psychophysiology 37(2000) 155–162
2
M.A Schier / Preprint Int. J. Psychophysiol 2000
1. Introduction The ability of an individual to control and operate a motor vehicle depends upon a wide variety of cognitive skills. While the cognitive processes associated with driving emphasise contributions of planning, memory, psychomotor control, and visual-spatial abilities, all depend on the central role of attention (Shinar 1993). Therefore the assessment of attentional abilities may prove useful in the assessment of driving abilities. One problem with attentional tests is that they cannot be used simultaneously with driving as they interfere with the underlying process and distort the fundamental task of driving, when used in the typical divided attention paradigm (for example Matthews et al. 1996). Waller (1992) described the need for using new technologies in transportation research on representative samples of users, particularly in the human factor aspect of highway research. One method of introducing new technology is to make use of simulations and simulators, which have several advantages to research. The benefits of simulators include safety, exposure to high-risk, low likelihood events, and cost-effectiveness. The impact of simulator learning on performance in the real situation has been evaluated, and is standard practice in the aviation, and space industries for example. However, the applications of simulators in driver training have not yet been established as ‘standard’practice. Simulators have been used to examine participants’responses to such parameters as road curvature (Land & Horwood 1995). Evidence to date indicates that driving simulators provide an adequate representation of the real world (Hoskovec & Stikar 1992; Hoffmann & Mortimer 1994; Schiff et al 1994; Galley 1993). Vehicle simulators can be separated into two distinct types. 1) Expensive, fully instrumented vehicles, for example the UMTRI vehicle (Sweet & Green 1993). 2) Inexpensive computer game based simulators exist, such as “Indycar”
(Micropose 1998), “Need for Speed” (Electronic Arts 1998), and “Grand Prix 2” (Micropose 1998). Originally this type of system was considered only as a game. With the advent of more sophisticated graphics cards and monitors, the simulation has become more realistic, especially when coupled with specialist hardware of realistic and familiar vehicle controls. An example of one such system, is the “Hyperstimulator” (Crooke 1998). This type of simulation software has a ‘replay’feature that can be exploited to review performance. The systems are of reasonable integrity, and some International Formula-One teams were reported making use of them for driver practice, to allow drivers to 'learn' circuits, and consolidate attentional factors associated with the driving task (Crooke 1998).
2. EEG measures One method with demonstrated validity for measurement of attention uses brain electrical activity (EEG) measures. The EEG is known to be responsive to changes in state, particularly the alpha activity (for example, Butler & Glass 1987). Alpha activity has been sometimes referred to as ‘idling’ activity, and the reduction in alpha activity as a result of attentional activity is referred to as ‘alpha blocking’(Schwartz et al 1989). Alpha activity and beta activity appear to have an inverse relationship, so that beta activity will increase when processing demands increase (Butler & Glass, 1987; Papanicalaou, et al., 1986). In addition some researchers found that alpha activity reflects changes in external events (such as sensory information), while beta activity reflects internal events (such as mental manipulation tasks) (Ray & Cole 1985). This may be simplifying matters greatly, although it is in agreement with the in/attention model of Shaw (1996). The time-varying characteristics of the EEG have been related to temporal attentional events in participants. This dynamic study of alpha activity
3
M.A Schier / Preprint Int. J. Psychophysiol 2000
in response to the changing external environment is referred to as Event-Related Desynchronization or ERD (Pfurtscheller & Aranibar 1977). More recent work by the Austrian group relate synchronous brain alpha activity to deactivated cortical areas (Kalcher & Pfurtscheller 1995; Pfurtscheller et al 1996). The presence or absence of phase-locked events and their relationship to ERD and Event Related Synchronization (ERS) is explored in depth by Kalcher & Pfurtscheller (1995). The ERD is related to the Steady State Visually Evoked Potential (SSVEP), and shows similarities in the type of activity and events that reduce the SSVEP and alpha activity (Schier & Silberstein1992; Silberstein et al 1990). When the ERD is reduced, the SSVEP is also reduced. The ERD is one measure applicable to the assessment of attention.
3. Measures of Driver Performance
to measure without interfering with the task at hand. However, measurable and reliable electrophysiological correlates of attention exist and are the subject of this study. 4. Materials and Methods 4.1Participants Two male participants, aged 26 and 31 participated in the study. They were both righthanded and had used the simulator apparatus before the recording session. Participants had a practice drive for approximately 5 minutes after the recording electrodes were connected, and before the demonstration began. Driver 1 was a competitor in the current Australian Rally Championship. 4.2 Equipment
Other parameters that have been recorded during driving studies can be divided into three types— physiological, behavioural, and vehicular. Physiological measures include eye movements as measured by the electro-oculogram (EOG), electroencephalogram (EEG), heart rate, and respiration. Behavioural measures include, inventories of behaviour (such as, sleepiness scales, personality scales), and reaction times. Vehicular measures include steering wheel position, lane position, speed, and others. Studies have usually examined one or two of these measures (Brown 1994; Sweet & Green 1993; Brookhuis & de Waard 1993). These measures provide complementary information, and only one large study in the USA involved recording about 10000 hours of all three types of data from 80 truck drivers in California (Miller 1995). The current study will examine aspects of all three parameters in a simulation environment, with particular emphasis on the EEG information in the alpha band (8–13 Hz). In summary, driving imposes significant demands on attentional processes that are difficult
A realistic, simulator environment (with a steering wheel, foot pedals, and gear lever) was used in conjunction with a computerised driving task (‘Need For Speed’, Electronic Arts 1998). To provide near identical conditions for both participants, the driving environment was limited to one vehicle on a closed circuit (Autumn Valley), using the same car (Ferrari 512TR, with a manual gear change). The video screen presented an 'incar' view to the driver. The video screen was 1 metre from the driver, and subtended an angle of approximately 18 degrees horizontal and 13 degrees vertically. 4.2 Driving task Participants completed six laps of the circuit, and EEG recording took place during laps 2, and 5. After the recordings took place, the playback mode was used, and participants' EEG was again recorded during laps 2 and 5 (they sat with their
4
M.A Schier / Preprint Int. J. Psychophysiol 2000
hands on the steering wheel and observed the race from the same 'in-car' view). Laps 2 and 5 were chosen as they were the furthest apart without being the initial or the final laps. Lap 1 was not used, as it was a new driving environment, and driving style may not have been established. Lap 6 was not used, as the subject knew that it was the final lap, and may have altered driving style accordingly. The only behavioural measure was the time for lap completion and was obtained from the Need for Speed software, and tabulated for both drivers. 4.4 EEG recording EEG was recorded from 4 scalp sites F3, F4 (frontal) and P3, P4 (parietal). Recordings were made with respect to linked earlobes (A1+A2), and forehead as ground for the differential recording. The EEG was amplified 50,000 times, band-passed filtered between 0.5 and 50 Hz, and sampled at a rate of 200 Hz to an IBM compatible computer using specially written software in the ‘Labview’ environment (National Instruments Corporation 1994). Ninety seconds of data from 4 channels were recorded in each instance. Subsequent EEG frequency and power spectral analysis were carried out using the ‘DAOS’ environment (Laboratory Software Associates 1986). 4.5 EEG analysis Power spectra were computed after applying a Hanning window to data segments of 1.28 seconds. From the power spectra the area under the curve was computed for the band alpha (8-13 Hz) for each channel and segment. These values were normalised with respect to the area under the curve between 0.5-26 Hz for each channel and segment, and labelled relative power. The power spectra were averaged and integrated over a period of 5.12 s to produce dynamic values of alpha for each channel and segment. Relative rather than absolute alpha values were chosen to facilitate comparison of data between participants.
5. Results Electrode impedances were measured below 6 kΩ for all four channels, F3, F4, P3 & P4 in both participants. For the first driver, the replay data for lap 5 were not available, due to loss of reference electrode. For this reason the only replay data presented for both drivers are from lap 2. Table 1 shows alpha activity as a function of electrode location for the driving condition Table 1 Mean relative alpha activity for the driving and replay tasks for both driversa Location
Driving lap 2
Driving lap 5
Replay lap 2
Driver 1 F3
0.34
0.37
0.52
F4
0.36
0.37
0.73
P3
0.34
0.29
0.33
P4
0.38
0.32
0.34
Lap time
1:28
1:23
—
F3
0.24
0.30
0.20
F4
0.24
0.31
0.37
P3
0.24
0.25
0.25
P4
0.28
0.32
0.32
Lap time
1:20
1:18
—
Driver 2
The table also contains the behavioural data (times for both driving laps). a
An increase in alpha activity should be interpreted as less attentional activity, and a decrease as more attentional activity, as the alpha indicates 'idling' activity, (or is indexing spare capacity, Schwartz et al. 1989). The behavioural data indicated an improvement in lap times for both drivers. The trend is less clear from the mean level of alpha activity during each task.
5
M.A Schier / Preprint Int. J. Psychophysiol 2000 1.2
1
0.8
0.6
0.4
0.2
0 1
3
5
7
9
11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65
1.2
1
0.8
0.6
0.4
0.2
0 1
3
5
7
9
11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65
Fig. 1 Relative alpha activity plotted against time (or epoch number). The diamonds (t) correspond to driving lap 2, the squares (§ ) to driving lap 5, and the triangles (5) correspond to the replay of lap 2 respectively. The upper plot is for driver 1 taken from electrode F4, while the lower is from driver 2 at the same electrode site.
Figure 1 illustrates the nature of the alpha activity from both drivers during the three tasks (lap 2, lap 5 and replay 2) from electrode F4. An
inherent variability exists for the alpha activity during each of these tasks, with the replay task showing the greatest variability. From Figure 1,
6
M.A Schier / Preprint Int. J. Psychophysiol 2000
mean levels of the driving tasks appear similar for driver 1, while the replay task has markedly elevated levels from about 25 seconds (epoch 19) onwards. Driver 2 shows similar, but stronger separation between driving on lap 2 and 5, with greater activity in lap 5. The replay tasks also showed an elevation from the driving task but not too the same extent as for driver 1.
*
0.4 0.35 0.3 0.25
*
0.2 0.15 0.1 0.05
*
*
0
*
*
-0.05 -0.1
drive F3
drive F4
drive P3
drive P4
P4) x 70 (epochs) repeated measures ANOVA. For driver 1, MANOVA indicated a significant main effect of driving condition (F(2,64)=182.7; p
