Towards Successful Multi-user Brain-Computer Interface (BCI)

Gaming: Analysis of the EEG Signatures and Connectivity

Finda D. Putri, Hao Ding, Abdullah Garcia and Aleksandra Vuckovic

Center of Rehabilitation Engineering, University of Glasgow, Glasgow, U.K.

Keywords: EEG-based BCI, Multi-user BCI, BCI Gaming, Competitive Gaming, Social Interaction.

Abstract: The information related to the impact of multi-user BCI on cortical activity is still relatively limited. This

ongoing study performed a competitive multi-user BCI gaming that is based on alpha band operant

conditioning and explored the brain activity and connectivity during the most, and the least successful gaming

runs. Ten healthy adults were involved in three days of gaming experiments in pairs. Multi-channel paired t-

test found a significant decrease (p<0.05) of absolute alpha power in the frontal left hemisphere channels in

the dominant players during the most successful gaming compared to the baseline of the same group. This

decrease is associated with the frontal alpha asymmetry (FAA) that occurred in the leading players.

Connectivity estimation via partial directed coherence (PDC) was also performed, showing the deactivation

of brain networks during the successful gaming of the dominant players compared to their baseline which

might indicate the “networks switching” mechanism from resting state to a more-demanding cognitive task.

Different baseline connectivity patterns were also found in the group of dominant players compared to the

group of non-dominant players, suggesting the possibility of using baseline connectivity information as a

predictor of gaming performance.

1 INTRODUCTION

The development of BCI technology has increased

the interest of BCI application for entertainment

purpose, commonly formed as BCI games. Gaming is

a highly stimulating activity that induces different

kinds of cognitive responses, making it very

challenging, yet very appealing for BCI application.

More advanced BCI-gaming technology initiated

the emerging of multi-user BCI games. The general

requirement of a multi-user BCI has been described

as the involvement of two or more users with

integrated brain activity to a BCI application (Nijholt,

2015; Nijholt & Gürkök, 2013). Social interaction

tasks (i.e. cooperation and competition) are ideally

implemented in multi-user BCI paradigms (Bonnet,

Lotte & Lécuyer, 2013; Gürkök et al., 2013; Nijholt

& Gürkök, 2013), as it has been widely used in the

classic video gaming.

Multi-user BCI development is a complex

process. Several factors composed of technical

challenges like the BCI architecture design and the

classification accuracy to behavioral factors like the

effect of social interaction on BCI performance

require special attention. Multiple studies around

interactive multi-player BCI gaming have been

performed, mainly to test the BCI classification

accuracy using different types of control, e.g. Steady-

State Evoked Potential (SSVEP) (Cruz et al., 2017;

Gürkök et al., 2013), P300 (Korczowski et al., 2016;

Korczowski, Congedo & Jutten, 2015), motor

imagery (MI) (Bonnet et al., 2013), and the

combination of different paradigms such as SSVEP/

P300 with alpha power (Mühl et al., 2009). Some of

these studies have also reported the impact of gaming

interaction on the quality of BCI performance. For

example, Bonnet et al. using BrainArena, an MI-

based multi-user BCI game, which was presented as

a simple ball game, found that social interaction is not

necessarily compensating the quality of BCI

performance. Additionally, compared to the single

player setup, they reported that the users prefer multi-

player gaming due to the fun and motivational factors

(Bonnet et al., 2013).

However, contradicting results were found by

another study that investigated the different types of

game control (BCI control and classic mouse control)

and their implications on co-experience during a

collaborative BCI game called “Mind The Sheep!”

(Gürkök et al., 2013). They found that when using

BCI control, co-experience was reduced by

Putri, F., Ding, H., Garcia, A. and Vuckovic, A.

Towards Successful Multi-user Brain-Computer Interface (BCI) Gaming: Analysis of the EEG Signatures and Connectivity.

DOI: 10.5220/0008201400590065

In Proceedings of the 3rd International Conference on Computer-Human Interaction Research and Applications (CHIRA 2019), pages 59-65

ISBN: 978-989-758-376-6

collaborative interaction due to the users’ concern of

losing control if they were paying too much attention

to collaboration.

In the field of multi-user BCI gaming, the amount

of information related to the neurological impact of

social interaction is still relatively limited. A study by

Toppi et al. has explored the cortical changes and

effective connectivity in the brain of pairs of pilots

during a flight scenario inside the flight simulator

(Toppi et al., 2016). They found that in theta band (3-

7 Hz), there is a significant influence (p<0.05)

between the flight phase factor (take-off, cruising,

and landing) on connections’ density and efficiency,

and the involvement of frontal networks of all the

pilots. They also reported a significant increase

(p<0.05) of theta power in the frontal and parieto-

occipital areas of the co-pilots during take-off, and in

the frontal area of both pilots during landing. A

significant increase of parietal alpha power was also

found only in co-pilots during landing (the captain led

take-off and the co-pilot led the landing due to a

deliberate electrical failure applied to the captain’s

instrumentation). Although physical control was

applied instead of BCI control, this study has

successfully demonstrated the impact of interaction

and the role changes on the cortical activity during a

cooperative task that requires a high-degree of

concentration, which can be a useful reference for

interactive BCI gaming development, especially

when applying high concentration task.

The aim of the current study is to analyse the

electrophysiological changes of the brain during a

competitive multi-user BCI gaming that is based on

the alpha band, non-verbalised operant conditioning.

We analysed the cortical changes by measuring

relative alpha power during gaming and resting states.

Neuronal connectivity was estimated in both gaming

and resting states, specifically during the most and

least successful gaming runs. The following sections

of this paper will be organised as: the materials and

methods, the results, the discussion, and the

conclusion.

2 MATERIALS AND METHODS

2.1 Experimental Setup

Ten healthy able-bodied adults (mean age 26.9±4.14,

6 females and 4 males) participated in EEG

experiments where they were sorted into five pairs.

They signed the written consent form prior to the

experiments. Ethical permission was granted by the

University’s College Ethical Committee.

The application consisted of three main

components: MATLAB functions (MATLAB 2015a,

The Mathworks, Inc., USA), a Simulink model and a

JAVA (version 1.8.0) Graphical User Interface

(GUI). MATLAB worked as an entrance system for

the application and served as the connector and

controller of the Simulink model and JAVA GUI.

The JAVA GUI shown in Figure 1 displays two

bars at each side of the screen and a seesaw in

between. Each bar represents the fluctuation of the

percentage relative alpha power (RA) to the power of

wider frequency band of 2-30 Hz, with a moving

average window of 0.5 second provided from the

electrode Pz in both players. Scoring was achieved

when one player managed to increase the power

≥10% than the other to make their side of seesaw tilt

down, and hold it for at least 1 second. The bar

changes colour from blue to green whenever players

gain 1 point. Prior to gaming experiment, baseline RA

was measured from each player to set individual

thresholds, which were later used to calculate a

normalising coefficient (NC). NC was applied to the

input signal generated by the player with a higher

threshold in the pair. This was done in order to

equalise the initial conditions. Furthermore, these

coefficients were acquired by dividing the RA of the

player with the lower threshold (RA

Low

) by that of the

player with the higher threshold (RA

High

), explained

as follows:

NC = RA

Low

/ RA

High

(1)

This approach is expected to help the non-dominant

player to maintain their control over their bar even if

their opponent has significantly higher baseline alpha

power.

EEG signal was recorded by a g.USBamp (g.tec

medical engineering GmbH., Austria) amplifier. The

EEG electrodes arrangement was set following the

standardised 10-20 EEG electrode placement system

(Homan, Herman & Purdy, 1987). The impedance

was kept below 5 kΩ. Linked ear reference was used

and FCz was used as the ground. Sampling frequency

was set to 256 Hz. Online band-pass filter was set

between 0.5 and 60 Hz (and a notch filter at 50 Hz)

using 5

order infinite impulse response (IIR) digital

Butterworth filter within the g.USBamp.

During the EEG experiments, two players were

seated next to each other in front of one screen. They

were instructed to compete with each other by

increasing the power bar located on their side of the

screen and to ‘push down’ the seesaw such that it

would be heavier towards their side.

Each pair performed three experimental sessions

on three separate days, where each session consisted

CHIRA 2019 - 3rd International Conference on Computer-Human Interaction Research and Applications

of six sub-sessions (5 minutes each). Pre- and post-

gaming eyes-open baseline were recorded in a relaxed

state for 2 minutes in every session. For the first two

sessions, EEG was recorded from only Pz electrode,

and for the third session, 32 and 16 electrodes were

used by the dominant (D) and the non-dominant (ND)

group of players respectively, where the dominance

was decided based on the highest average scoring

performance in the previous two sessions.

Figure 1: Competitive gaming interface. Users have to

increase their power bar ≥10% than the other for at least 1

second to score 1 point.

2.2 Data Analysis

All EEG data were analysed using EEGLAB (UC San

Diego, SCCN, USA) toolbox in MATLAB. All EEG

data were band-pass filtered between 2 and 30 Hz.

For single-channel data, visual inspection and manual

noise removal were performed, and for multi-channel

data, independent component decomposition was

performed by using an Independent Component

Analysis (ICA) algorithm implemented in EEGLAB

for further noise removal.

The average power spectrum analysis was

estimated by using Welch’s method (Welch, 1967),

with a 50% overlap of 4 second long windows.

Individual RA at Pz was analysed at a group and

individual level. Spatial distribution of power in all

conditions was estimated based on multi-channel

EEG recording. Paired t-test analysis was applied to

multi-channel EEG data to test statistical significance

in absolute power between conditions.

PDC, a multivariate measurement of directional

causality in the frequency domain, was calculated to

estimate brain connectivity. PDC from i to j can be

defined as:

𝜋

𝑖𝑗

(

𝑓

)

𝐴

𝑖𝑗

(𝑓)

√

𝑎

𝑗

∗

(

𝑓

)

𝑎

𝑗

(𝑓)

(2)

Where 𝐴

𝑖𝑗

(𝑓) is the i,j-th element of A(f), a matrix of

frequency domain transformed model coefficient, and

𝑎

𝑗

(𝑓) is the j-th column of matrix A(f) (Baccalá &

Sameshima, 2001). PDC values were measured from

10 representative electrodes and the significant PDC

values were estimated by using asymptotic statistic (p

< 0.05) and False Discovery Rate (FDR) correction (p

< 0.05) was applied for multiple comparisons.

3 RESULTS

Scoring results were used to measure the BCI

performance of the users. From all three sessions (in

a total of 30 gaming sub-sessions per session for all 5

pairs), on average, group D won all the time in the

first and second session and won only 80% (24 sub-

sessions) in the third session (group ND won 6 sub-

sessions). Results in Figure 2 show the average

percentage of individual gaming RA along with the

average scoring performance, from all sessions. The

bars with an asterisk represent the players with higher

baseline RA, where NC was applied to their RA

during gaming (these players received feedback of

their normalised RA instead of their real RA). Our

results suggest that higher baseline RA does not

always reflect better performance. Players with lower

baseline RA can still win the game. It also shows that

daily adjustment of the baseline RA was necessary, as

in some pairs, different players had larger/smaller

baseline RA on different days.

In multi-channel data, two gaming conditions

were selected based on the highest/lowest scores of

the last session. Table 1 shows the scores of the

highest and the lowest scoring gaming sub-sessions in

both groups during the last experimental session.

Based on this measure, the multi-channel analysis

was grouped into the highest scoring D, lowest

scoring D, highest scoring ND, and lowest scoring

ND. We then categorised highest scoring D as “the

most successful gaming” and the lowest scoring ND

as “the least successful gaming”.

Paired t-test (p<0.05) of the multi-channel data

across subjects (FDR correction applied) found

significant decrease of absolute alpha power only in

the most successful gaming compared to their

baseline, specifically in the frontal electrodes of the

left hemisphere (FP1, AF3, FC5, and FC3), as seen in

Figure 3. There is no statistical significance found in

the theta and beta bands in all conditions and groups,

showing the selectivity of EEG power modulation in

the alpha band only. Although there is no significant

increase found on the training electrode, Figure 3

Towards Successful Multi-user Brain-Computer Interface (BCI) Gaming: Analysis of the EEG Signatures and Connectivity

shows that spatially source of high alpha was reduced

around Pz during successful gaming.

Figure 4 shows the estimated PDC in the alpha

band, which reflects the connections among networks

during baseline of both groups, and during the most

successful gaming condition, highest scoring D (red

square), and its counterpart- the other player on that

particular session, and during the least successful

gaming, lowest scoring ND (blue square), and its

counterpart. Ten electrodes covering four brain

cortical areas were chosen as the representative nodes

(i.e., F3, Fz, F4, C3, Cz, C4, P3, Pz, P4 and Oz). Our

observations found higher connectivity during

baseline in group D (players with higher baseline

RA), which are indicated by the higher estimated

PDC values and the more complex connected

networks, in contrast, the baseline of the group with

lower RA (Figure 4 top right) shows relatively lower

estimated PDC values and less connected networks.

Table 1: The highest scoring (Hi) and the lowest scoring

(Lo) gaming sub-sessions (SS) in both groups.

Pair

Score

149

134

Score

114

Score

105

Score

139

105

During the most successful gaming (red square),

our results show that the connectivity is decreasing

(from baseline state) in terms of the PDC values and

the number of network connections, and the

remaining networks from baseline are found around

the frontal electrodes. The ND counterpart of the most

successful gaming shows increased connectivity

around frontal and central areas, specifically in the

left hemisphere (F3 ↔ Fz and C3 ↔ Cz) with an

emerging connection from C3 to P3 compared to their

baseline state. Similar to their counterparts, occipital

connectivity is decreasing compared to baseline. This

ND group is just the counterpart of the most

successful gaming, not necessarily depicting the

highest scoring condition for the group ND.

For the least successful gaming (blue square),

compared to their baseline, the number of

connections found are low, with the only strong

remaining connection found from F3 to Fz. The D

counterpart of the least successful gaming, shows

higher PDC values in the parieto-occipital

connectivity, compared to their baseline and the

highest scoring D. In contrast to the most successful

gaming, group D of this particular condition shows

more complex connectivity, reflecting that this is not

their successful gaming performance, despite still

dominating group ND during gaming.

Figure 2: (a) Mean and standard deviation of individual

gaming RA (%) at Pz from all sessions. The bars with

asterisks show normalised RA, which used as the feedback

to the players with higher baseline RA, and the bars without

asterisks show real RA. (b) Mean and standard deviation of

scores from all users from all sub-sessions.

Figure 3: Paired t-test analysis shows a significant decrease

of alpha in some frontal electrodes during the highest

scoring gaming sub-sessions compared to baseline before

gaming for group D.

CHIRA 2019 - 3rd International Conference on Computer-Human Interaction Research and Applications

Figure 4: PDC estimation results show connectivity in the

alpha band in ten electrodes during baseline (top) and the

highest scoring D (red square) and its counterpart

representing the most successful gaming (middle), and the

lowest scoring ND (blue square) and its counterpart

(bottom).

4 DISCUSSION

The present study aimed to examine the effect of

competitive gaming on brain cortical activity and

neuronal connectivity. The results from the average

of the individual gaming RA and the average scoring

suggest that higher average RA (both in baseline and

during gaming) does not directly produce higher

scores. We found a case where the winning player had

a lower average RA during gaming. One reason for

this is that during gaming, players with higher

baseline RA received feedback on normalised, i.e.

lower RA than their original RA. In order to score,

timing is essential, as someone can only score if they

manage to hold their bar higher for at least 1 second.

Thus, in this case, the losing player was not able to

sustain high alpha power for a relatively long period,

though on average, their RA might be high. BCI

control requires skill and training is necessary to help

the users obtaining and maintaining that skill. In this

setup, the players were not only expected just to

increase their RA, but also to control its duration and

to play based on the opponent’s feedback which is

changing over time.

Our multi-channel spectral analysis shows a

significant decrease of frontal alpha power during the

most successful gaming performed by the dominant

players. It has been known that naturally, alpha power

tends to decrease during task-engagement (Bazanova

& Vernon, 2014) and frontal alpha suppression has

been reported during interactive synchronized finger-

tapping task of two subjects, particularly stronger in

the leading subjects, reflecting their higher cognitive

investment during the synchronized action

(Konvalinka et al., 2014). However, in our results,

significant activation only occurred in one side of the

hemisphere, indicating that this suppression might be

related to the frontal alpha asymmetry (FAA) which

is defined as the different frontal alpha activity

between hemispheres (Davidson et al., 1990). FAA

has been associated with the motivation of

approaching and withdrawing behavior, to be

specific, if the left hemisphere is more activated, then

it is associated with approaching behavior rather than

withdrawing behavior (Coan & Allen, 2004), thus our

results might reflect the motivation to be engaged in

the task by the winning players. Furthermore, left

hemisphere activation in a social setting has been

associated with unsocial and anti-social behavior

(Hecht, 2014), which is in line with our competitive

setting, where the players were expected to play

against each other.

Connectivity estimation results in the alpha band

during successful gaming have demonstrated the

connectivity pattern changes, which consisted of the

deactivation of several network connections from

resting state to gaming performance. Previous

observations of the changing connectivity pattern

between resting state and cognitive task, have

indicated the involvement of three different brain

networks such as the default mode network (DMN),

the central executive network (CEN), and the salience

network (SN) (Goulden et al., 2014; Seeley et al.,

2007; Sridharan, Levitin & Menon, 2008). The DMN

has been defined as a group of networks which is

found to be more active when the task-engagement is

absent whereas the CEN is a group of network that is

activated when the brain is engaged in a specific

mental task (Greicius et al., 2003; Raichle et al., 2001;

Seeley et al., 2007). The SN, which comprised of the

ventrolateral prefrontal cortex, fronto-insular cortex,

and anterior cingulate cortex, is known as the

mediator network that helps the network switching

between task-free and task-engagement states,

between the DMN and the CEN (Goulden et al.,

2014). An alpha neurofeedback training was reported

Towards Successful Multi-user Brain-Computer Interface (BCI) Gaming: Analysis of the EEG Signatures and Connectivity

increasing the SN connectivity, where this increased

connectivity was also found to be negatively

correlated with mind-wandering task and resting

alpha rhythm (Ros et al., 2013), two conditions which

activate the DMN (Neuner et al., 2014; Simon &

Engström, 2015).

Our connectivity results also show different

connectivity patterns during baseline in different

groups. This difference might reflect the possibility

of using resting state connectivity to predict gaming

performance, where the similar idea has been

proposed in predicting the neurofeedback training

response in individuals with anxiety (Scheinost et al.,

2014).

In order to obtain more detailed connectivity

information, especially regarding the pattern changes

between the three major networks, more specific and

larger number of electrodes should be chosen for

connectivity analysis. Larger number of subjects and

different range of frequency bands are also required

to explore the different impact of interactive BCI

gaming in different frequency bands. An equal

number of electrodes for both players should be used

for the last gaming session to avoid bias by the

more/less dominant players.

5 CONCLUSIONS

Our study introduced a multi-user competitive BCI

game that is based on alpha operant conditioning. We

reported the preliminary results of the cortical

changes and connectivity from the most successful

gaming. Further development of this study will

include more participants and other social interaction

settings (i.e., collaborative), in order to explore the

brain activity and connectivity during the different

interaction tasks.

ACKNOWLEDGEMENTS

This study is funded by Indonesian Endowment Fund

for Education by the Ministry of Finance, Republic of

Indonesia.

REFERENCES

Baccalá, L. A., & Sameshima, K. J. B. c. (2001). Partial

directed coherence: a new concept in neural structure

determination. 84(6), 463-474.

Bazanova, O., & Vernon, D. (2014). Interpreting EEG

alpha activity. Neuroscience & Biobehavioral Reviews,

44, 94-110.

Bonnet, L., Lotte, F., & Lécuyer, A. (2013). Two brains,

one game: design and evaluation of a multiuser BCI

video game based on motor imagery. IEEE

Transactions on Computational Intelligence, 5(2), 185-

198.

Coan, J. A., & Allen, J. J. (2004). Frontal EEG asymmetry

as a moderator and mediator of emotion. Biological

psychology, 67(1-2), 7-50.

Cruz, I., Moreira, C., Poel, M., Ferreira, H., & Nijholt, A.

(2017). Kessel Run-A Cooperative Multiplayer SSVEP

BCI Game. Paper presented at the International

Conference on Intelligent Technologies for Interactive

Entertainment.

Davidson, R. J., Ekman, P., Saron, C. D., Senulis, J. A., &

Friesen, W. V. (1990). Approach-withdrawal and

cerebral asymmetry: emotional expression and brain

physiology: I. Journal of Personality Social

Psychology, 58(2), 330.

Goulden, N., Khusnulina, A., Davis, N. J., Bracewell, R.

M., Bokde, A. L., McNulty, J. P., & Mullins, P. G.

(2014). The salience network is responsible for

switching between the default mode network and the

central executive network: replication from DCM.

Neuroimage, 99, 180-190.

Greicius, M. D., Krasnow, B., Reiss, A. L., & Menon, V.

(2003). Functional connectivity in the resting brain: a

network analysis of the default mode hypothesis.

Proceedings of the National Academy of Sciences,

100(1), 253-258.

Gürkök, H., Nijholt, A., Poel, M., & Obbink, M. (2013).

Evaluating a multi-player brain–computer interface

game: Challenge versus co-experience. Entertainment

Computing, 4(3), 195-203.

Hecht, D. (2014). Cerebral lateralization of pro-and anti-

social tendencies. Experimental Neurobiology, 23(1),

1-27.

Homan, R. W., Herman, J., & Purdy, P. (1987). Cerebral

location of international 10-20 system electrode

placement. Electroencephalogr Clin Neurophysiol,

66(4), 376-382.

Konvalinka, I., Bauer, M., Stahlhut, C., Hansen, L. K.,

Roepstorff, A., & Frith, C. D. (2014). Frontal alpha

oscillations distinguish leaders from followers:

multivariate decoding of mutually interacting brains.

Neuroimage, 94, 79-88.

Korczowski, L., Barachant, A., Andreev, A., Jutten, C., &

Congedo, M. (2016). " Brain Invaders 2": an open

source Plug & Play multi-user BCI videogame. Paper

presented at the 6th International Brain-Computer

Interface Meeting (BCI Meeting 2016).

Korczowski, L., Congedo, M., & Jutten, C. (2015). Single-

trial classification of multi-user P300-based Brain-

Computer Interface using riemannian geometry. Paper

presented at the 2015 37th annual international

conference of the IEEE engineering in medicine and

biology society (EMBC).

CHIRA 2019 - 3rd International Conference on Computer-Human Interaction Research and Applications

Mühl, C., Gürkök, H., Bos, D. P.-O., Thurlings, M. E.,

Scherffig, L., Duvinage, M., . . . Heylen, D. (2009).

Bacteria Hunt: A multimodal, multiparadigm BCI

game. Paper presented at the Fifth International

Summer Workshop on Multimodal Interfaces.

Neuner, I., Arrubla, J., Werner, C. J., Hitz, K., Boers, F.,

Kawohl, W., & Shah, N. J. (2014). The default mode

network and EEG regional spectral power: a

simultaneous fMRI-EEG study. PLoS One, 9(2),

e88214.

Nijholt, A. (2015). Competing and collaborating brains:

multi-brain computer interfacing. In Brain-Computer

Interfaces (pp. 313-335): Springer.

Nijholt, A., & Gürkök, H. (2013). Multi-brain games:

cooperation and competition. Paper presented at the

International Conference on Universal Access in

Human-Computer Interaction.

Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W.

J., Gusnard, D. A., & Shulman, G. L. (2001). A default

mode of brain function. Proceedings of the National

Academy of Sciences, 98(2), 676-682.

Ros, T., Théberge, J., Frewen, P. A., Kluetsch, R.,

Densmore, M., Calhoun, V. D., & Lanius, R. A. (2013).

Mind over chatter: plastic up-regulation of the fMRI

salience network directly after EEG neurofeedback.

Neuroimage, 65, 324-335.

Scheinost, D., Stoica, T., Wasylink, S., Gruner, P., Saksa,

J., Pittenger, C., & Hampson, M. (2014). Resting state

functional connectivity predicts neurofeedback

response. Frontiers in Behavioral Neuroscience, 8,

338.

Seeley, W. W., Menon, V., Schatzberg, A. F., Keller, J.,

Glover, G. H., Kenna, H., . . . Greicius, M. D. (2007).

Dissociable Intrinsic Connectivity Networks for

Salience Processing and Executive Control. Journal of

Neuroscience, 27(9), 2349-2356. doi:10.1523/

JNEUROSCI.5587-06.2007 %J The Journal of

Neuroscience

Simon, R., & Engström, M. (2015). The default mode

network as a biomarker for monitoring the therapeutic

effects of meditation. Frontiers in Psychology, 6, 776.

Sridharan, D., Levitin, D. J., & Menon, V. (2008). A critical

role for the right fronto-insular cortex in switching

between central-executive and default-mode networks.

Proceedings of the National Academy of Sciences,

105(34), 12569-12574.

Toppi, J., Borghini, G., Petti, M., He, E. J., De Giusti, V.,

He, B., . . . Babiloni, F. (2016). Investigating

cooperative behavior in ecological settings: an EEG

hyperscanning study. PLoS One, 11(4), e0154236.

Welch, P. (1967). The use of fast Fourier transform for the

estimation of power spectra: A method based on time

averaging over short, modified periodograms. IEEE

Transactions on Audio and Electroacoustics, 15(2), 70-

73. doi:10.1109/TAU.1967.1161901

Towards Successful Multi-user Brain-Computer Interface (BCI) Gaming: Analysis of the EEG Signatures and Connectivity