An SSVEP-Actuated Brain Computer Interface Using Phase-Tagged ...

Annals of Biomedical Engineering, Vol. 38, No. 7, July 2010 (© 2010) pp. 2383–2397 DOI: 10.1007/s10439-010-9964-y

An SSVEP-Actuated Brain Computer Interface Using Phase-Tagged Flickering Sequences: A Cursor System PO-LEI LEE,1,2,3 JYUN-JIE SIE,1 YU-JU LIU,1 CHI-HSUN WU,1 MING-HUAN LEE,1 CHIH-HUNG SHU,2,4,5 PO-HUNG LI,2,6 CHIA-WEI SUN,7,8 and KUO-KAI SHYU1 1 Department of Electrical Engineering, National Central University, No. 300, Jhongda Rd., Jhongli City, Taoyuan County 32001, Taiwan; 2Integrated Brain Research Laboratory, Taipei General Veterans Hospital, Taipei, Taiwan; 3Institute of Brain Science, National Yang-Ming University, Taipei, Taiwan; 4Department of Otolaryngology, Taipei Veterans General Hospital, Taipei, Taiwan; 5National Yang-Ming University School of Medicine, Taipei, Taiwan; 6Department of Otolaryngology, Cheng-Hsin Rehabilitation Medical Center, Taipei, Taiwan; 7Biophotonics Interdisciplinary Research Center, National Yang-Ming University, Taipei, Taiwan; and 8Institute of Biophotonics, National Yang-Ming University, Taipei, Taiwan

(Received 23 November 2008; accepted 10 February 2010; published online 23 February 2010) Associate Editor Berj L. Bardakjian oversaw the review of this article.

Keywords—Brain–computer interface (BCI), Steady-state visual evoked potential (SSVEP), Electroencephalography (EEG), Phase-tagged flickering sequence.

Abstract—This study presents a new steady-state visual evoked potential (SSVEP)-based brain computer interface (BCI). SSVEPs, induced by phase-tagged flashes in eight light emitting diodes (LEDs), were used to control four cursor movements (up, right, down, and left) and four button functions (on, off, right-, and left-clicks) on a screen menu. EEG signals were measured by one EEG electrode placed at Oz position, referring to the international EEG 10-20 system. Since SSVEPs are time-locked and phase-locked to the onsets of SSVEP flashes, EEG signals were bandpass-filtered and segmented into epochs, and then averaged across a number of epochs to sharpen the recorded SSVEPs. Phase lags between the measured SSVEPs and a reference SSVEP were measured, and targets were recognized based on these phase lags. The current design used eight LEDs to flicker at 31.25 Hz with 45° phase margin between any two adjacent SSVEP flickers. The SSVEP responses were filtered within 29.25–33.25 Hz and then averaged over 60 epochs. Owing to the utilization of high-frequency flickers, the induced SSVEPs were away from low-frequency noises, 60 Hz electricity noise, and eye movement artifacts. As a consequence, we achieved a simple architecture that did not require eye movement monitoring or other artifact detection and removal. The high-frequency design also achieved a flicker fusion effect for better visualization. Seven subjects were recruited in this study to sequentially input a command sequence, consisting of a sequence of eight cursor functions, repeated three times. The accuracy and information transfer rate (mean ± SD) over the seven subjects were 93.14 ± 5.73% and 28.29 ± 12.19 bits/min, respectively. The proposed system can provide a reliable channel for severely disabled patients to communicate with external environments.

INTRODUCTION Patients suffering from severe motor disabilities, such as amyotrophic lateral scleroses (ALS), severe cerebral palsy, head trauma, multiple sclerosis, and muscular dystrophies, are incapable of communicating with external environments.58,59 Several research groups have dedicated themselves to developing novel techniques, which allow users to control external devices or express their intentions independent of peripheral neuromuscular functions. Among those proposed solutions, one promising technique, called brain computer interface (BCI), was developed to help patients communicate with external environments by means of recording specific brain signals induced from elaborately designed tasks, and then translating the measured brain signals into communication or control signals. Over the past few decades, several BCI systems have been designed based on various kinds of brain signals. Pfurtscheller et al. measured sensorimotor Mu rhythms during subjects’ imagery movements and achieved a high recognition rate of 90%.40 Birbaumer et al. developed a thought translation device (TTD) to measure slow cortical potentials (SCPs) for a binary selection task.3 Donchin et al. implemented an alphabet typing system based on visually-induced P300.11 Mason and Birch designed an asynchronous detector to control a binary switch based on detecting motorrelated potentials (MRPs) filtered within 1–4 Hz.33

Address correspondence to Po-Lei Lee, Department of Electrical Engineering, National Central University, No. 300, Jhongda Rd., Jhongli City, Taoyuan County 32001, Taiwan. Electronic mail: [email protected]

2383 0090-6964/10/0700-2383/0

© 2010 Biomedical Engineering Society

2384

LEE et al.

However, due to the inherent nature of brain signals utilized in the aforementioned systems, the information transfer rates (ITRs) are relatively low, on the order of 5–25 bits/min at best.58,59 Accordingly, some research groups are engaging themselves in developing exogenous BCIs for faster ITRs, such as visual evoked potential (VEP)-based BCI systems.6,25–27 VEP has been reported as a stable response in EEG measurements, which has been used to monitor anesthesia level during surgery,41,55 to diagnose prechiasmal and retrochiasmal lesions,5,23,32,57 to indicate intracranial pressure induced by a head injury,36 and to alarm brain death.42,54 Due to the neural connections and interactions of the route from the retina to the primary visual cortex, VEP is time-locked and phaselocked to the onset of visual stimulus.20,46–48,52 VEP has the characteristic of being dominated by the response induced from the visual stimulus located at the subject’s central visual field,2,4,50,51 which is called “cortical magnification”.50,51 The cortical magnification refers to the numbers of neurons (and hence the amount of cortex) responsible for processing a stimulus of a specific size. Due to the greater numbers of neurons representing the foveal visual field, the cortical surface associated with foveal stimuli is proportionately larger, resulting in a larger VEP. This electrophysiological characteristic permits a user’s gazed target to be recognized by checking the temporal or spectral relation, such as latency or frequency response, between the recorded VEP and the flickering sequences for driving the flicker at gazed position.17,25–27 At least four different VEP-based BCI systems have been developed. Sutter developed a brain response interface (BRI) by measuring fast multifocal visual evoked potentials (FMFVEPs) induced from a pseudorandom sequence.50 The measured brain signals were correlated with a stereotypical designed “response template” to detect the gaze target by checking the latency of maximum correlation.50 Lee et al. utilized mutually independent flickering sequences to induce onset and offset flash visual evoked potentials (FVEPs).25–27 Onset and offset FVEPs in central visual field were enhanced by means of a time-locked average process. Kelly et al. implemented attention-regulated SSVEP-based BCI which permits a users’ SSVEP amplitudes to be regulated by their attention levels.21,53 Cheng et al. used multifrequency flickers to induce user’s steady-state visual evoked potentials (SSVEPs) and the gazed-target was detected by checking the spectral peaks on the estimated spectrum.7,56 Nevertheless, the aforementioned systems utilized multiple frequencies or random flickers as visual stimuli. The stimulation frequencies in those systems were low and the induced VEPs typically existed in low frequency

ranges (95%) with 60 or more epochs being averaged for gazed-target detections. Accordingly, 60 epochs have been adopted to compromise the computation and accuracy for faster feedback in our application studies in this current system. Table 1 lists the detected tref and the θdiff obtained from the seven volunteers in our control study, in which the tref was the latency of maximum amplitude in SSVEPrefs obtained from gazing at LED1 and the θdiff was the phase difference between the detected phase lag (θd) in SSVEPgazes and expected phase delays (θi) (i.e., θdiff = θd − θi, i = 2, 3,…, 8). The SSVEPgazes were obtained from an average of 60 nonoverlapped epochs. The obtained θdiff (mean ± SD) (248 SSVEPgazes in each subject) were −1.46° ± 4.09°, 3.3° ± 13.95°,

2.7° ± 14.98°, 0.8° ± 10.84°, −2.0° ± 12.40°, −4.7° ± 16.2°, and −6.4° ± 8.61°, respectively, and the average and standard deviation of θdiff was −0.42° ± 11.69°. Figure 8 illustrates one example of the raster plot of SSVEPgazes induced from subject I when trying to produce the cursor command sequence ‘ ⇦BL⇧⇩BR⇨ ’ in the application study. Each SSVEPgaze was obtained by an average of the latest 60 consecutive epochs relative to the time point of gazedtarget detection. The phase lag, θd, in each SSVEPgaze was detected, and SSVEPgazes were expressed in degrees of phase lags between 0° and 360° based on their phase lags (θd). The SSVEPgazes obtained from 3 to 41 s were vertically aligned with their amplitudes presented in color scale. The first 2-s were skipped due to the insufficient epoch number (N24 bits/min) with little or no training. Our system detects phase lag by finding the time difference between the waveform peaks of SSVEPgaze and SSVEPref, which differs from the correlation-based technique used in FMFVEP-based system.50 Furthermore, the FMFVEP-based system presumes an identical response of VEP across all trials and uses it as a template for the correlation process. Such an assumption might be too stringent, since the latencies, peak amplitude, waveforms, etc., of human VEPs can vary from trial-to-trial, and using such a stereotypical template for correlation may result in incorrect detection when using a short data length.20,24,26,34 Third, the phase information in the induced SSVEP is a stable index for BCI control. Figure 7 shows this stability of the phase lag over time for two subjects, and Table 1 shows the detected phase lag was as expected across subjects and targets. The phase detection approach enables dividing a full cycle (360°) into several phase regions and assigning each phase region with a corresponding cursor function. Very few studies have investigated phase information in SSVEP.

2393

Strasburger used a grating stimuli with different spatial frequencies to study the consistency of phase angles in SSVEPs,49 and concluded the SSVEP phase as a remarkably stable index even when the SSVEP amplitude is very low. Ding et al. studied the effects of attention modulations on SSVEPs and found that both the phaselocking index (PHI) and the power of SSVEPs were enhanced, especially in theta and alpha bands, when the subjects were concentrating on the gazed flicker.10 Eriksen et al. interpreted this attention-regulation phenomenon as a “spotlight” effect,14 which enhances the cortical representation of stimulus presented in attended regions. The attention–regulation effect has also been demonstrated in a number of EEG studies,8,16,18,30,60,62 and neuro-physiological studies.13,29,35,43 Fourth, our system employs high-frequency flickers for visual stimulations and enables all LEDs flashing at the same frequency. Since studies have reported a messy display achieved by SSVEP flickers as a key factor leading to user fatigue,56 SSVEP flickers with a high-frequency design can cause flicker fusion effect to attain a more comfortable visualization.9 A highfrequency visual stimulation (>30 Hz) has the advantage of avoiding the induced SSVEP from being interfered with low-frequency environmental noise and some brain rhythms as well. Compared to other SSVEPbased BCI using multiple frequencies for visual stimulation, flickering frequencies around alpha and beta bands should be excluded and high-frequency flickers (>20 Hz) were seldom used due to lack of an effective procedure for high-frequency SSVEP extractions.7,28,56 This study carefully selected the flickering frequency at 31.25 Hz, reported as the stimulation frequency for inducing the largest SSVEP amplitudes within the middle frequency range.56 The chosen frequency is far enough away from the ranges of alpha and beta rhythms, EOG artifacts, and electricity noise (50 or 60 Hz) noise to allow robust phase detection.24,31 In our application studies, the epoch-average process was performed every 0.1 s, and one target was recognized as valid only if same target was detected in 15 successive gazed-target detections. Every SSVEPgaze was obtained by an average of the latest 60 epochs anchored to the time point for gazed-target detection, and each valid cursor command required the same target to be detected for 15 successive times. Therefore, the theoretically fastest CTI for producing a valid input was 60 (cycles)/31.25 (cycles/s) + 0.1 (s/detection) 9 (15 − 1) (detections) = 3.32 s. We chose 0.1 s as time step for gazed-target detection, because it was the hardware limitation of AD card in our experiment. The minimum time interval for data transmission between AD card (NI-PCI 6071E, National Instruments Co.) and our LabView program was 0.1 s, and, therefore, shorter time interval for data transmission

2394

LEE et al.

FIGURE 9. Test of the relation between the confirmation number and accuracy in three subjects. With 5, 10, 15, and 20 confirmation times, the accuracy of valid command outputs were 65.96 ± 7.13%, 69.6 ± 6.82, 94.96 ± 5.63, and 94.77 ± 2.14%, respectively. It can be seen that 15 confirmations can produce satisfactory results (94.96 ± 5.63%) and detection accuracy in 20 confirmations produced similar results but with smaller standard variations.

was infeasible in our current setup. Regarding the 15 confirmation for producing a valid cursor command, it was determined empirically since we hardly found that one target was recognized in 15 successive times but he/ she was not gazing at a flickering target. We repeated the application study using different number of confirmation from three subjects. Figure 9 shows the relations between the confirmation number and accuracy. With 5, 10, 15, and 20 confirmation times, the accuracy of valid command outputs were 65.96 ± 7.13%, 69.6 ± 6.82, 94.96 ± 5.63, and 94.77 ± 2.14%, respectively. It can be seen that 15 confirmations can produce satisfactory results (94.96 ± 5.63%) and detection accuracy in 20 confirmations produced similar results but with smaller standard variations. Nevertheless, the rules for optimal performance will be studied in future works. The proposed system identifies users’ gazed-targets by checking the phase lag between SSVEPgaze and SSVEPref, in which high SNR in the acquired SSVEP is necessary to ensure satisfactory phase detection. In other words, the standard deviation of the phase differences (θdiff) between detected phase lags (θd) and the expected phase delays (θi) should be low so that the system has high accuracy. To investigate the feasibility and accuracy of the proposed system, we designed a control study with two purposes: one was to check how the predicted phase difference varied from actual and the other was to examine the feasibility of the system in distinguishing eight different targets. For the first purpose of the control study, we detected phase lags (θd) in SSVEPgazes with average of 60 nonoverlapped epochs and the phase difference (θdiff) over the seven subjects which was −0.42° ± 11.69°. If phase difference (θdiff) were normally distributed, we anticipated that

95% of phase lags lay within two-times the standard deviation (2 9 ±11.69° = ±23.38°). With phase margin of ±23.38°, the available target number was 7.69 targets (360°/(2 9 23.38°) = 7.69). Taking the nearest integer, the proposed system had eight different targets which resulted in the phase margin as ±22.5°, and the predicted accuracy was 94.56% with the given phase difference (θdiff) of −0.42° ± 11.69°. The predicted result appeared to agree with the accuracy of 93.14 ± 5.73% performed by experienced subjects in Table 2(b). The second purpose of the control study examined the accuracies of the system (see Fig. 6) and tested how the difference of the phase lags in SSVEPs (see Fig. 7) varied from predicted phase delays, with different epoch number being applied in the epochaverage process. Owing to the high accuracies (>95%) being achieved with 60 or more epochs for epochaveraging process, the 60 epochs were chosen to achieve faster feedback, i.e., shorter command transfer interval (CTI), with satisfactory high accuracy. Although the average over a number of epochs was considered as an intuitive way to suppress SSVEPunrelated noise in real time, the mean ITR and accuracy of the proposed system can be further improved in the following two ways. First, advanced signal processing techniques, such as independent component analysis (ICA),20,24,26,34,61 and the wavelet-based method,45 can be applied to extract the SSVEPs with higher SNR so that fewer epochs might be possible in the epoch-average process for suppressing peripheral visual responses. Second, classifiers, such as artificial neural network (ANN),15,39 support vector machine (SVM),37 and linear discriminate analysis (LDA)11,19 could be helpful to decide optimized phase margins among SSVEPgazes induced from different flickers. Cheng et al. designed an optimal bipolar measurement with the use of ICA for their SSVEP-based BCI and have improved the ITR from 27.15 to 43 bits/min. Meinicke et al. adopted SVM and increased the mean ITR of their P300-based BCI from 12 to 50.5 bits/min.37 It is worthy to notice that the subject’s familiarity with the experiment and their attention level may affect detection rates. In the application study I, comparing the overall results between the experienced (subject I) and the naı¨ ve groups (subject II to subject VII) (see Table 2a), the experienced subject had better accuracy (100% vs. 80.75%) and higher ITRs (35.22 vs. 22.46 bits/min) than the naı¨ ve group. In Table 2(b), previous training showed improvement on subjects’ execution performance, which achieved better accuracies (98% vs. 91.2%) and higher ITRs (41.21 vs. 23.10 bits/min) in the experienced group (subject I and subject II) than the inexperienced group (subject III to subject VII). We observed the experienced subject has

SSVEP-Based BCI Suing Phase Encoded Flickering Sequences

better concentration on the visual stimulus than the naı¨ ve or inexperienced subjects who were distracted occasionally by surrounding nontarget stimuli. For example, subject IV in application study I produced two wrong commands ‘ ’ and ‘ ’, which were located close to the intended target ‘⇧’, and subject VI reported that he was incautiously distracted by the wrong flicker ‘ ’ for three consecutive times when he was trying to gaze at flicker ‘⇧’. Another example in the application study II shows that subject IV produced five wrong commands ‘⇨’, which were located close to the intended target ‘ ’, and reported that he was incautiously distracted several times by the wrong flicker ‘⇨’ when he was trying to gaze at the desired target ‘ ’. Having previous training, the accuracy and ITR in the application study II were improved from 83.49% to 93.14% and from 24.28 to 28.29 bits/ min, respectively, compared to the results in the application study I. Accordingly, although no training is usually asserted in such VEP-based BCI system,7,28,50,51,58 we still found that previous experience may improve the subject’s performance in this study. Additionally, one may notice that CTI in the application task II was longer than that in the application task I (6.03 vs. 4.83 s), even though all the participants in the application task II had more experience in performing this SSVEP task. Since subject’s attention level has been demonstrated as a factor to regulate SSVEP,10,14,62 loss of attention caused by mental fatigue, distraction, etc., may lower the signal quality of SSVEP. In our application task II, subjects kept high attention to pass the rigorous rule (15 successive confirmations) for producing valid command outputs in a long experiment time (Ttotal = 157.86 ± 63.56). Discontinuity of high concentration in 3.32 s (15 successive gazed-target confirmations) in each command production will prolong the required total experiment time (Ttotal) for command sequence production. Although we have recognized that both task familiarity and attention are two important issues in performing this SSVEP-based BCI, subject’s attention level is difficulty to be quantified in our current experiment design. Future studies are required to study the influence of attention on SSVEP performance. Though both the VEP-based BCI system and the popular eye tracker systems are associated with eye motions, the proposed SSVEP-based BCI system has several superior features to those eye tracker systems. First, eye trackers detect retro-reflectivity of reference points, e.g., reflection from the pupil center and the cornea of a stationary light source, based on image analysis. More than two cameras are needed to capture stereovision images for continuous tracking of reference points, and, therefore, only minor head movement is allowed to prevent miscalculations.12 Besides, a

2395

stationary environment is usually required to prevent the influence of glints from surrounding false objects.12 In contrast, our system utilizes stable flickers with a specific frequency to induce subject’s SSVEPs. A simple averaging process is sufficient to effectively suppress the effects from ambient task-unrelated glints. Second, available visual angles for eye tracking systems typically operate within ±45° to guarantee that the boundaries of the iris or cornea can be well captured. Compared to our system, the SSVEP-based BCI can be well operated as long as the user’s eye perceives light from visual stimulator. Third, eye saccade velocities can be up to 700°/s within a duration as short as 20 ms,6 and thereby most video-based eye trackers are equipped with high-speed video capture systems (>250 Hz).12 In contrast, the system demand for VEP-based BCI is low and comprises only an Oz EEG channel and a microprocessor for data processing.7,21,25,26,38,50,51,56 CONCLUSIONS This work proposes a SSVEP-based BCI using phase-tagged flickering sequence to produce cursor commands for communication purposes. Subjects shift their gazes at different LED flickers and phase information of the induced SSVEP is extracted for recognizing the gazed-targets. The salient features of the proposed system are: (1) SSVEP is a very stable and reliable neuro-electric signal to be detected; (2) phasetagged flickering sequences are adopted and only one flickering frequency is used; (3) SSVEP-unrelated noise can be removed by simply applying bandpass filtering and an epoch-average process; (4) high-frequency flickers are used so that the induced SSVEPs can avoid interferences from low-frequency noises; (5) the highfrequency design achieves a more comfortable visualization. Our current system enables encoding eight distinct phases on a flickering frequency. The proposed system can be further extended by reducing the phase margin or utilizing more flickering frequencies to increase the number of available commands. The proposed system provides an efficient and reliable channel for the neuromuscular disabled to communicate with external environments.

ACKNOWLEDGMENTS This study was funded by the National Central University, National Science Council (95-2314-B-075118, 96-2628-E-008-070-MY3, 96-2221-E-008-122-MY3, 96-2221-E-010-003-MY3, 96-2221-E-008-115-MY3,96-2752B-010-008-PAE), and Veterans General Hospital University

LEE et al.

2396

System of Taiwan Joint Research Program (VGHUST96-P415, VGHUST97-P3-11, VGHUST98-P3-09, VGHUST99-P313). We thank Prof. Yu-Te Wu for his contribution in manuscript preparation.

REFERENCES 1

Basar, E. Brain functions and oscillation. In: CrossModality Experiments on the Cat Brain, edited by E. Basar, T. Demiralp, M. Schurmann, and C. BasarEroglu. Berlin: Springer-Verlag, 1999, pp. 27–59. 2 Baseler, H. A., E. E. Sutter, S. A. Klein, and T. Carney. The topography of visual evoked response properties across the visual field. Electroencephalogr. Clin. Neurophysiol. 90:65–81, 1994. 3 Birbaumer, N., H. Flor, N. Ghanayim, T. Hinterberger, I. Iverson, E. Taub, B. Kotchoubey, A. Kubler, and J. Perelmouter. A spelling device for the paralyzed. Nature 398:297–298, 1999. 4 Brown, B., and M. Z. Yu. Variation of topographic visually evoked potentials across the visual field. Ophthal. Physl. Opt. 17:25–31, 1997. 5 Carlin, L., E. S. Roach, A. Riela, E. Spudis, and W. T. McLean. Juvenile metachromatic leukodystrophy: evoked potentials and computed tomography. Ann. Neurol. 13:105–106, 1983. 6 Carpenter, R. H. S. Movements of the Eyes (2nd ed.). London, England: Pion, 1988. 7 Cheng, M., X. Gao, S. Gao, and D. Xu. Design and implementation of a brain–computer interface with high transfer rates. IEEE Trans. Biomed. Eng. 49:1181–1186, 2002. 8 Clark, V. P., and S. A. Hillyard. Spatial selective attention affects early extrastriate but not striate components of the visual evoked potential. J. Cogn. Neurosci. 8:387–402, 1996. 9 Cornsweet, T. N. Visual Perception. New York: Academic, 1970. 10 Ding, J., G. Sperling, and R. Srinivasan. Attentional modulation of SSVEP power depends on the network tagged by the flicker frequency. Cerebral Cortex 16:1016– 1029, 2006. 11 Donchin, E., K. M. Spencer, and R. Wilesinghe. The mental prosthesis: assessing the speed of a P300-based braincomputer interface. IEEE Trans. Rehabil. Eng. 8:174– 179, 2000. 12 Duchowski, A. T. Eye tracking methodology: theory and practice. In: Eye Tracking Technologies. London, England: Springer Publishers, 2003, pp. 55–65. 13 Fries, P., J. H. Reynolds, A. E. Rorie, and R. Desimone. Modulation of oscillatory neuronal synchronization by selective visual attention. Science 291:1560–1563, 2001. 14 Eriksen, C. W., and J. D. St. James. Visual attention within and around the field of focal attention: a zoom lens model. Percept. Psychophys. 40:225–240, 1986. 15 Haselsteiner, E., and G. Pfurtscheller. Using time-dependent neural networks for EEG classification. IEEE Trans. Rehabil. Eng. 8:457–463, 2000. 16 Heinze, H. J., G. R. Mangun, W. Burchert, H. Hinrichs, M. Scholz, T. F. Munte, A. Gos, M. Scherg, S. Johannes, H. Hundeshagen, M. S. Gazzaniga, and S. A. Hillyard. Combined spatial and temporal imaging of brain activity

during visual selective attention in human. Nature 372:543– 546, 1994. 17 Herrmann, C. S. Human EEG responses to 1–100 Hz flicker: resonance phenomena in visual cortex and their potential correlation to cognitive phenomena. Exp. Brain Res. 137:346–353, 2001. 18 Hillyard, S. A., and L. Anllo-Vento. Event-related brain potentials in the study of visual selective attention. Proc. Natl. Acad. Sci. USA 95:781–787, 1998. 19 Hinterberger, T., A. Kubler, J. Kaiser, N. Neumann, and N. Birbaumer. A brain–computer interface (BCI) for the locked-in: comparison of different EEG classifications for thought translation device. Clin. Neurophosiol. 114:416–425, 2003. 20 Jung, T. P., S. Makeig, M. Westerfield, J. Townsend, E. Courchesne, and J. T. Sejnowski. Analysis and visualization of single-trial event-related potentials. Hum. Brain Mapp. 14:166–185, 2001. 21 Kelly, S. P., E. C. Lalor, R. B. Reilly, and J. J. Foxe. Visual spatial attention tracking using high density SSVEP data for independent brain–computer communication. IEEE Trans. Neural Syst. Rehabil. Eng. 13:172–178, 2005. 22 Krishnaveni, V., S. Jayaraman, S. Aravind, V. Hariharasudhan, and K. Ramadoss. Automatic identification and removal of ocular artifacts from EEG using wavelet transform. Meas. Sci. Rev. 6:45–57, 2006. 23 Kriss, A., W. M. Carroll, L. D. Blumhardt, and A. M. Halliday. Pattern and flash evoked potential changes in toxic (nutritional) optic neuropathy. Adv. Neurol. 32:11–19, 1982. 24 Lee, P. L., Y. T. Wu, L. F. Chen, Y. S. Chen, C. M. Cheng, T. C. Yeh, L. T. Ho, M. S. Chang, and J. C. Hsieh. ICA based spatiotemporal approach for single-trial analysis of post-movement MEG beta synchronization. Neuroimage 20:2010–2030, 2003. 25 Lee, P. L., C. H. Wu, Y. T. Wu, L. F. Chen, T. C. Yeh, and J. C. Hsieh. Visual evoked potential (VEP)—actuated brain computer interface: a brain-actuated cursor system. Electron. Lett. 21:832–834, 2005. 26 Lee, P. L., J. C. Hsieh, C. H. Wu, K. K. Shyu, S. S. Chen, T. C. Yeh, and Y. T. Wu. The brain computer interface using flash visual evoked potential and independent component analysis. Ann. Biomed. Eng. 34:1641–1654, 2006. 27 Lee, P. L., J. C. Hsieh, C. H. Wu, K. K. Shyu, and Y. T. Wu. Brain computer interface using flash onset and offset visual evoked potentials. Clin. Neurophysiol. 119:605–616, 2008. 28 Lin, Z., C. Zhang, W. Wu, and X. Gao. Correlation analysis for SSVEP-based BCIs. IEEE Trans. Biomed. Eng. 54:1172–1176, 2007. 29 Luck, S. J., L. Chellazzi, S. A. Hillyard, and R. Desimone. Neural mechanisms of spatial selective attention in areas V1, V2, and V4 macaque visual cortex. J. Neurophysiol. 77:24–42, 1997. 30 Mangun, G. R., and S. A. Hillyard. Spatial gradients of visual attention: behavioral and electrophysiological evidence. Electroencephalogr. Clin. Neurophysiol. 75:417–428, 1988. 31 Manoilov, P. EEG eye-blinking artefacts power spectrum analysis. CompSysTech IIIA:1–5, 2006. 32 Markand, O. N., B. P. Garg, W. E. DeMyer, and C. Warren. Brain stem auditory, visual and somatosensory evoked potentials in leukodystrophies. Electroencephalogr. Clin. Neurophysiol. 54:39–48, 1982. 33 Mason, S. G., and G. E. Birch. A brain-controlled switch for asynchronous control applications. IEEE Trans. Biomeed. Eng. 47:1297–1307, 2000.

SSVEP-Based BCI Suing Phase Encoded Flickering Sequences 34

McKeown, M. J., S. Makeig, G. G. Brown, T. P. Jung, S. S. Kindermann, A. J. Bell, and T. J. Sejnowski. Analysis of fMRI data by blind separation into independent spatial components. Hum. Brain Mapp. 6:160–188, 1998. 35 McMains, S. A., and D. C. Somers. Multiple spotlights of attentional selection in human visual cortex. Neuron 42:677–686, 2004. 36 McSherry, J. W., C. L. Walters, and J. D. Horbar. Acute visual evoked potential changes in hydrocephalus. Electroencephalogr. Clin. Neurophysiol. 53:331–333, 1982. 37 Meinicke, P., M. Kaper, F. Hoppe, M. Heumann, and H. Ritter. Improving transfer rates in brain computer interfacing: a case study. Adv. Neural. Inf. Proc. Syst. 15:1131–1138, 2003. 38 Middendorf, M., G. McMillan, G. Calhoun, and K. S. Jones. Brain–computer interface based on the steady-state visual-evoked response. IEEE Trans. Neural Syst. Rehabil. Eng. 8:211–214, 2000. 39 Palaniappan, R., R. Paramesran, S. Nishida, and N. Saiwaki. A new brain–computer interface using fuzzy ARTMAP. IEEE Trans. Neural. Syst. Rehabil. 10:140–148, 2002. 40 Pfurtscheller, G., C. Neuper, C. Guger, W. Harkam, H. Ramoser, A. Schlogl, B. Obermaier, and M. Pregenzer. Current trends in Graz brain–computer interface (BCI) research. IEEE Trans. Rehabil. Eng. 8:216–219, 2000. 41 Raitta, C., U. Karhunene, A. M. Seppalainen, and M. Naukkarinen. Changes in the electroretinogram and visual evoked potentials during general anaesthesia. Albrecht von Graefes Arch. Klin. Exp. Ophthalmol. 211:139–144, 1979. 42 Reilly, E. L., C. Kondo, J. A. Brunberg, and D. B. Doty. Visual evoked potentials during hypothermia and prolonged circulatory arrest. Electroencephalogr. Clin. Neurophysiol. 45:100–106, 1978. 43 Reynolds, J. H., and L. Chelazzi. Attentional modulation of visual processing. Annu. Rev. Neurosci. 27:611–647, 2004. 44 Schurmann, M., and E. Basar. Topography of alpha and theta responses upon auditory and visual sitmuli in humans. Biol. Cybern. 72:161–174, 2004. 45 Sivakumar, R., B. Hema, P. Karir, and N. Nithyaklyani. Denosing of transient VEP signals using wavelet transform. J. Eng. Appl. Sci. 1:242–247, 2006. 46 Spehlmann, R. Evoked potential primer. In: Electrode Placements and Combinations for Full-Field and HalfField VEPs. Stoneham, MA: Butterworth Publishers, 1985, pp. 103–109. 47 Spehlmann, R. Evoked potential primer. In: The Transient VEP to Diffuse Light Simuli, edited by K. E. Misulis, and T. Fakhoury. Stoneham: Butterworth Publishers, 1985, pp. 135–142. 48 Spehlmann, R. Evoked potential primer. In: VEPs to Other Stimuli, edited by K. E. Misulis, and T. Fakhoury. Stoneham: Butterworth Publishers, 1985, pp. 144–158.

49

2397

Strasburger, H., W. Wolfgang, and I. Rentschler. Amplitude and phase characteristics of the steady-state visual evoked potential. Appl. Opt. 27:1069–1088, 1988. 50 Sutter, E. E. The brain response interface: communication through visually-induced electrical brain responses. J. Microcomput. Appl. 15:31–45, 1992. 51 Sutter, E. E., and D. Tran. The field topography of ERG components in Man—I. The photopic luminance response. Vision Res. 32:433–446, 1992. 52 Tang, A. C., B. A. Pearlmutter, N. A. Malaszenko, and D. B. Phung. Independent components of magnetoencephalography: single-trial response onset times. Neuroimage 17:1773–1789, 2002. 53 Trejo, L. J., R. Rosipal, and B. Matthews. Brain–computer interfaces for 1-D and 2-D cursor control: designs using volitional control of the EEG spectrum or steady-state visual evoked potentials. IEEE Trans. Neurol. Syst. Rehabil. 14:225–229, 2006. 54 Trojaborg, W., and E. O. Jorgensen. Evoked cortical potentials in patients with “isoelectric” EEGs. Electroencephalogr. Clin. Neurophysiol. 35:301–309, 1973. 55 Uhl, R. R., K. C. Squires, D. L. Bruce, and A. Starr. Effect of halothane anesthesia on the human cortical visual evoked response. Anesthesiology 53:273–276, 1980. 56 Wang, Y., R. Wang, X. Gao, B. Hong, and X. Gao. A practical VEP-based brain–computer interface. IEEE Trans. Neural Syst. Rehabil. Eng. 14:234–239, 2006. 57 Wilson, W. B. Visual-evoked response differentiation of ischemic optic neuritis from the optic neuritis of multiple sclerosis. Am. J. Ophthal. 86:530–535, 1978. 58 Wolpaw, J. R., N. Birbaumer, W. J. Heetderks, D. J. McFarland, P. H. Peckham, G. Schalk, E. Donchin, L. A. Quatrano, C. J. Robinson, and T. M. Vaughan. Brain– computer interface technology: a review of the first international meeting. IEEE Trans. Neural Syst. Rehabil. Eng. 8:164–173, 2000. 59 Wolpaw, J. R., N. Birbaumer, D. J. McFarland, G. Pfurtscheller, and T. M. Vaughan. Brain–computer interfaces for communication and control. Clin. Neurophysiol. 113:767–791, 2002. 60 Worden, M. S., J. J. Foxe, N. Wang, and G. V. Simpson. Anticipatory biasing of visuospatial attention indexed by retinotopically specific alpha-band electroencephalography increases over occipital cortex. J Neurosci 20:RC63, 2000. 61 Wu, C. H., P. L. Lee, Y. T. Wu, and J. C. Hsieh. ICAbased analysis of movement-related modulation on beta activity of single-trial MEG measurement using spatial and temporal templates. J. Med. Biol. Eng. 28:155–159, 2008. 62 Yamaguchi, S., H. Tsuchiya, and S. Kobayashi. Electroencephalographic activity associated with shifts of visuospatial attention. Brain 117:553–562, 1994.