The Eects of Eye Trcking in VR Helmet on EEG Recordings Jessic D. Byliss nd Dn H. Bllrd The University of Rochester Computer Science Deprtment Rochester, New York 14627 Technicl Report 685 My 1998 Astrct Since sclp EEG recordings re mesured in microvolts, electricl signls my esily interfere during n experiment. As Spehlmnn discusses, such interference my e introduced through the lights in the recording room, nery television, or even computer monitor [Spehlmnn, 1991]. Thus, when we consider performing EEG/EP/ERP experiments within virtul relity helmet contining n eye trcker, electricl interference ecomes rel possiility. We tested the eects of wering VR4 virtul relity (VR) helmet contining n ISCAN eye trcker while sking sujects to do continuous performnce tsk. The results of this tsk were then nlyzed in the frequency domin nd compred to results from the sme experiment while looking t computer screen in two dierent environments. Results indicte tht in n environment with other computers, the verticl refresh from the ck of nery row of computer monitors dded more noise to the signl thn wering the VR helmet nd eye trcker. Even in n environment without other computers, the noise while wering the VR helmet nd eye trcker is not signicntly dierent from the noise otined while viewing lptop computer screen in the sme loction. This work ws supported y NIH/PHS reserch grnt P41 RR09283-03 nd ws fcilitted in prt y Ntionl Physicl Science Consortium Fellowship nd y stipend support from NASA Goddrd Spce Flight Center.
1 Introduction The use of virtul relity (VR) nd eye trcking during EEG experiments promises to e enecil. Virtul relity cn provide dynmic nd controllle experimentl environments nd eye trcking llows comprison etween cognitive ehvior s ssocited with eye movement nd through the EEG signl. While [Nelson et l., 1997] recorded EEG signls in virtul environment, the environment in question consisted of lrge screen in front of the suject. This screen posed less of prolem with noise thn we ssumed virtul relity helmet sitting on the hed of suject would. Since sclp EEG recordings re mesured in microvolts, electricl signls my esily interfere during n experiment. As discussed in [Spehlmnn, 1991], such interference my e introduced through the lights in the recording room, nery television, or even computer monitor. Thus, when we consider performing EEG/EP/ERP experiments within VR4 virtul relity helmet contining n ISCAN eye trcker, electricl interference ecomes rel possiility. Mny lgorithms exist for reducing known rtifcts from recording (see [d Silv et l., 1986], [Sdsivn nd Dutt, 1995], [Berg nd Scherg, 1994], nd [Jung et l., 1997] for smple of methods), ut in order to reduce the rtifct it is helpful to rst know wht kind of rtifct exists. We tested the eects of wering the VR helmet with the eye trcker while sking sujects to do continuous performnce tsk. The results of this tsk were then nlyzed in the frequency domin nd compred to results from the sme experiment while looking t two dierent kinds of computer screen in two dierent environments. In the \noisy" environment, other computers were present nd they were sent in the \quiet" environment. Results indicte tht the verticl refresh from the ck of computer monitor in nery row dded more noise to the signl thn wering the VR helmet nd eye trcker. Even in n environment without other computers, the noise while wering the VR helmet nd eye trcker is not signicntly dierent from the noise otined while viewing lptop computer screen in the sme loction. 2 Experiments Electrodes were plced on the sujects using the Interntionl 10-20 system with linked mstoid reference. Figure 1 shows the electrode plcement for oth sujects. Electrode impednce ws elow 20 kohms on ll chnnels nd the signl ws mplied using Grss mpliers. The nlog frequency rnge ws from 0.3 Hz to 300 Hz. While this is rther lrge rnge, our interest in exmining noise justies the rnge. A Neuro Scn Acquisition system ws then used to digitlly record the signls t rte of 1000 Hz. Discovered in the 1960's [Sutton et l., 1965], the P300 is n esily evoked positive wveform round 300 ms fter infrequent tsk relevnt stimuli. Using visul oddll prdigm, the tsk for oth sujects ws to click mouse utton upon seeing n \E" on the screen they were viewing nd to do nothing when seeing n \S", \O", \W", or \X". E's occurred 15% of the time. The Neuro Scn Stim visul continuous performnce tsk progrm ws used oth to sh letters on the screen nd to trigger the EEG cquisition 1
VEOG HEOG S1 Only S2 Only FZ S1 nd S2 C3 CZ PZ C4 P4 Figure 1: The electrode plcements for oth sujects using the Interntionl 10-20 system with linked mstoid reference. Computer Suject Computer Computer Computer Figure 2: The environmentl setup for the rst suject. Notice how the cks of the monitors in nery row fce the suject. system to record only signls occurring etween 100 ms efore the stimulus nd 1000 ms fter the stimulus ws presented. The interstimulus intervl ws 2 seconds. For the rst experiment, S1 viewed the stimulus from Hitchi SuperScn Elite 17 inch monitor. This monitor hs vrile verticl refresh, which ws 70 Hz t the time of the experiment nd the suject st pproximtely 1.5 feet wy from the monitor. The monitor ws locted in n environment with other computers (including the EEG signl recording computer) in the setup of Figure 2. On dierent dy, the VR helmet ws plced on S1's hed nd S1 repeted the originl tsk while viewing the hed-mounted disply inside the helmet with the eye trcker turned on nd in dierent room from the previous experiment where there were no other computers close to the suject. An verge of the signl from oth types of disply is shown in Figure 3. The verge frequency spectrum for ech type of disply over ll epochs is shown in Figure 4. As cn e seen from the results, dominnt 70 Hz frequency is seen in 2
- + Figure 3:. The verge of the rre (P300) trils recorded with electrode PZ from S1 while viewing 17 inch monitor in room with other computers.. The verge of the rre (P300) trils recorded with electrode PZ from S1 while wering the VR helmet. All signls were low pss ltered t 100 Hz (the usul low pss frequency for P300 experiments) efore verging. Notice tht. ppers noisier thn. due to the 70 Hz monitor verticl refresh coming from the ck of nery computer monitor. the environment with other computers. This ws due to the 70 Hz monitor verticl refresh coming from the ck of computer monitor pproximtely 2 feet from the suject's hed. While the prolem with the signl power of the nery monitor verticl refresh could not e denied from this experiment, severl other prolems with the experiment itself rose fter creful considertion. First, the two prts of the experiment were done on dierent dys. While the impednce t the electrode sights for oth dys ws elow 20 kohms, the impednces were proly not similr enough to e le to directly compre the signl power in the two prts of the experiment. After crefully reding the mnul for the helmet, it ws discovered tht the helmet signl ws sent t rte of 60 Hz due to use of the NTSC timing formt [Systems, 1994]. While recording during the rst experiment, the 60 Hz notch lter ws on nd thus potentil prolem with the VR helmet signl ws voided. An experiment using second suject xed these prolems: the notch lter remined o, oth prts of the experiment were done in room with no close computers nd the lights o, nd oth experiments were done in sequence on the sme dy. Suject S2 performed the sme tsk s S1, ut t rst viewed the tsk from the 12 inch screen of lptop computer. Lptop computers hve cklit screens nd it ws resoned tht there would e no possiility of verticl refresh prolem with this type of screen. During the sme recording session, the VR helmet with the eye trcker ws crefully plced on the suject's hed nd electrode impednces were rechecked. The imge signl from the 12 inch monitor ws converted to S-VHS formt nd piped into the VR helmet. A smple verge under oth viewing conditions is shown in Figure 5. A frequency nlysis of ll epochs is shown in Figure 6. The signl power in the frequency spectrum ws similr in oth prts of the experiment, lthough the VR helmet hd more signl power round 60-70 3
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Figure 4:. A frequency nlysis of the verge of the mgnitude of the power spectrum signl from ech EEG epoch for S1 while looking t 17 inch computer monitor. The 70 Hz signl is due to the verticl refresh from the ck of nery computer monitor.. A frequency nlysis of the verge of the mgnitude of the power spectrum signl from ech EEG epoch for S1 while wering the VR helmet nd eye trcker. The dierences in the low frequency signl power re most likely due to dierent quntity of eye movements etween sessions. Both grphs re shown t the sme scle nd re low pss ltered t 300 Hz. Hz. Due to the construction of the VR helmet, we found tht the nd of the helmet going round the hed hit ginst electrode loctions P3 nd P4. This is why electrode loction P4 ws recorded in ddition to the recorded loctions for S1. The verge power spectrum for this electrode loction is shown in Figure 7. Surprisingly, there isn't much of dierence etween the two prts of the experiment. Ech epoch of chnnel P4 ws crefully checked nd it ws found tht the electrode did not suddenly move on the hed during the time of the recording. 3 Discussion In this pper we hve shown tht the verticl refresh from the ck of nery monitor my contin more noise thn wering VR helmet with n eye trcker. The overll noise levels while viewing 12 inch lptop monitor nd while wering the helmet were comprle to one nother - indicting tht experiments done in the VR helmet while trcking suject's eye will not e more noisy thn sitting in front of lptop screen. The results of this experiment were mostly positive, ut must e tempered with cution in n ctul virtul environment. For instnce, the helmet nd presses ginst electrode loctions P3 nd P4. The suject remined still in this experiment, looking t sttic screen. In rel virtul relity environment, the suject will e expected to move his/her hed in order to see dierent 4
- + Figure 5:. The verge of rre (P300) trils recorded with electrode PZ from S2 while viewing 12 inch lptop screen.. The verge of rre (P300) trils recorded with electrode PZ from S2 while wering the VR helmet. These verges re much noisier looking thn S1's verges simply ecuse only 10 rre trils were verged for. nd. The signls were digitlly low pss ltered t 100 Hz (the usul nlog low pss frequency for P300 experiments) efore verging. x 10 4 x 10 4 7 4 6 3.5 5 3 2.5 4 2 3 0 10 20 30 40 50 60 70 80 90 100 1.5 0 10 20 30 40 50 60 70 80 90 100 2 1 1 0.5 0 0 10 20 30 40 50 60 70 80 90 100 0 0 10 20 30 40 50 60 70 80 90 100 Figure 6:. A frequency nlysis of the verge of the mgnitude of the power spectrum signl from ech EEG epoch for S2 while looking t the 12 inch screen on lptop computer.. A frequency nlysis of the verge of the mgnitude of the power spectrum signl from ech EEG epoch for S2 while wering the VR helmet nd eye trcker. Slightly more noise exists inside the VR helmet, ut the two frequency wveforms correlte t 0.79 (p<0.05) for the dt from 20-100 Hz (where most of the noise ws ssumed to lie). Both plots were clculted directly from the 300 Hz nlog low pss ltered dt. 5
Figure 7:. 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 A frequency nlysis of the verge of the mgnitude of the power spectrum signl from ech EEG epoch for S2 while looking t the 12 inch disply on lptop computer for electrode P4.. A frequency nlysis of the verge of the mgnitude of the power spectrum signl from ech EEG epoch for S2 while wering the VR helmet nd eye trcker for electrode P4. Notice tht there is little more noise overll while wering the helmet. This my e due to the electrode shifting little when the helmet ws plced on the hed. For oth. nd. the impednce fell elow 20 kohms. Both grphs re shown t the sme scle. prts of the world. This my cuse the P3 nd P4 electrodes to shift from ruing ginst the helmet if the helmet is not tight enough round the hed. In the future, recordings need to e checked for the possiility of shifting electrodes. 6
References [Berg nd Scherg, 1994] P. Berg nd M. Scherg, \A multiple source pproch to the correction of eye rtifcts," Electroenceph. Clin. Neurophysiol., 90(3):229{241, 1994. [d Silv et l., 1986] F.H. Lopes d Silv, W.Storm vn Leeuwen, nd A. Remond, editors, Hndook of Electroencephlogrphy nd Clinicl Neurophysiology: Volume 2, Clinicl Applictions of Computer Anlysis of EEG nd other Neurophysiologicl Signls, N. Elsevier Science Pulishers, Amsterdm, 1986. [Jung et l., 1997] T.P. Jung, C. Humphries, T. Lee, S. Mkeig, M.J. McKeown, V. Irgui, nd T.J. Sejnowski, \Extended ICA Removes Artifcts from Electroencephlogrphic Recordings," to Apper in Advnces in Neurl Informtion Processing Systems, 10, 1997. [Nelson et l., 1997] W.T. Nelson, L.J. Hettinger, J.A. Cunninghm, nd M.M. Roe, \Nvigting through virtul ight environments using rin-ody-ctuted control," Proceedings. IEEE 1997 Virtul Relity Annul Intern. Symposium, pges 30{37, 1997. [Sdsivn nd Dutt, 1995] P.K. Sdsivn nd D. Nryn Dutt, \Line interference cncelltion from corrupted EEG signls using modied liner phse FIR digitl lters," Proceedings of the First Regionl Conference, IEEE Engineering in Medicine nd Biology Society nd 14th Conference of the Biomedicl Engineering Society of Indi, pges 3/35{36, 1995. [Spehlmnn, 1991] R. Spehlmnn, Spehlmnn's EEG Primer, N. Elsevier Science Pulishers, Amsterdm, 1991. [Sutton et l., 1965] S. Sutton, M. Brren, J. Zulin, nd E. John, correltes of stimulus uncertinty," Science, 150:1187{1188, 1965. \Evoked potentil [Systems, 1994] Virtul Reserch Systems, User's Guide: VR4, Virtul Reserch Systems, Inc., Snt Clr, CA, 1994. 7