EEG AND HUMAN LOCOMOTION

Descending Commands and Sensory Feedback should be Disentangled

from Artifacts Thanks to New Experimental Protocols

Position Paper

Thierry Castermans

, Matthieu Duvinage

, Guy Cheron

and Thierry Dutoit

TCTS lab, Faculty of Engineering, Universit´e de Mons, 31 Boulevard Dolez, 7000, Mons, Belgium

LNMB lab, Universit´e Libre de Bruxelles, CP 168, 50 Av F. Roosevelt, Brussels, Belgium

Keywords:

EEG Analysis, Human Locomotion, Artifacts, Brain-computer Interfaces.

Abstract:

The main challenge when studying EEG signals related to human walk control comes from the fact that

signals of many different origins are mixed up. Indeed, descending commands from the brain are generated,

while ascending sensorimotor information coming from the feet is sent to the brain. In addition to the inherent

complexity of the human control mechanism, experimental investigation of the cerebral activity elicited during

walk is highly challenging: electrode movements are produced by movements of the head, but also by the

shocks undergone by the whole body at each step, which – albeit signiﬁcantly attenuated – are transmitted to

the head and degrade the quality of EEG signals. Recently, different EEG studies of human locomotion have

been published. These are based on different hypotheses and/or produce results that are contradictory. After

reviewing and describing the discrepancies between the different approaches, we propose new experimental

protocols which should help to solve important issues.

1 INTRODUCTION

Human locomotion is known to be based on a very

complex hierarchical system which includes several

control networks located at spinal and supraspinal

levels (Hanakawa, 2006). Basically, high-level mo-

tor commands are sent by the brain to a spinal net-

work composed of central pattern generators (CPG)

and, at the same time, each level of motor control re-

ceives peripheral sensory information (sensory feed-

back) which is used to modify the motor output at that

level.

The central pattern generators network consists

in coupled antagonist oscillators speciﬁcally dedi-

cated to extensor or ﬂexor muscles acting at the dif-

ferent joints. Their mechanism allows to generate

simple and coordinated rhythmic movements such as

those involved in steady walk (Kiehn, 2006). Ahead

of the CPG, supra-spinal networks (i.e. the brain-

stem, cerebellum and cortex) are of crucial impor-

tance in the control of walking. Indeed, as summa-

rized in (Presacco et al., 2011) and references therein,

signiﬁcant changes in motor and cognitive demands

(i.e. spatial attention) have been observed in the con-

text of bipedal walking in unknown or cluttered dy-

namic environments. Functional neuroimaging stud-

ies have shown that the primary motor cortex is re-

cruited during rhythmic foot or leg movements. Ad-

ditionally, the technique of functional near-infrared

spectroscopy (fNIRS) has allowed to detect involve-

ment of frontal, premotor and supplementary motor

areas during walking (Harada et al., 2009), (Suzuki

et al., 2008).

Electroencephalography (EEG) represents an in-

teresting complementary technique for investigating

neural processes governing walk. Indeed, while stan-

dard functional imaging is characterized by a good

spatial resolution, EEG is the only wearable and non-

invasive measurement technique which offers a tem-

poral resolution good enough in order to study the dy-

namics of brain. However, electrophysiologicalinves-

tigation of the cerebral activity elicited during walk

is highly challenging. Indeed, head and body move-

ments constitute an important source of mechanical

artifacts strongly affecting the EEG signals quality.

This explains why very few papers have been writ-

ten on the subject.

Recently, EEG studies of human locomotion have

309

Castermans T., Duvinage M., Cheron G. and Dutoit T..

EEG AND HUMAN LOCOMOTION - Descending Commands and Sensory Feedback should be Disentangled from Artifacts Thanks to New Experimental

Protocols Position Paper.

DOI: 10.5220/0003871403090314

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (BIOSIGNALS-2012), pages 309-314

ISBN: 978-989-8425-89-8

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

been undertaken but give incompatible results. In this

context, the objective of this paper is twofold: (1) re-

view and discuss the new available results, empha-

sizing contradictory aspects as well as discrepancies

with independent experimental results; (2) propose

new experimental protocols to resolve the ambigui-

ties.

2 EEG STUDIES OF HUMAN

LOCOMOTION

2.1 Static vs Dynamic Approaches

EEG signals are by essence noisy and difﬁcult to mea-

sure (a few microvolts only). This is due to the fact

that each EEG scalp electrode is only able to measure

the combination (superposition) of the electrical po-

tentials generated by thousands of neurons, which are

weakened and smeared by the volume conduction ef-

fect of the skull (Nunez et al., 1997). Moreover, EEG

signals may be affected by different kinds of artifacts

generated either by extracerebral physiological activ-

ity (blinks, eye movements, muscle or cardiac activ-

ity), by interference with power line, or by record-

ing electrodes and equipment. On top of this, typical

artifacts related to gait further degrade EEG signals

quality when the measurements are made in ambula-

tory conditions (see for instance (Castermans et al.,

2011) for a review). The combination of these mul-

tiple effects renders EEG study of human locomotion

extremely complex.

Consequently, the main strategy generally used to

overcome these experimental difﬁculties consists in

focusing on simpliﬁed foot or leg movements which

imply common cerebral processes with gait. In these

experimental protocols, subjects are mainly static and

produce only limited lower limb movements. A

strong advantage of this approach is of course that

motion artifacts are drastically limited. In this case,

however, the full neural activity related to walk is

not available and, for instance, cerebral processes

involved in posture and balance control are miss-

ing. Recording EEG signals of subjects walking on a

treadmill include of course all these aspects but then

requires a powerful analysis technique to discriminate

the different artifact contributions from the real corti-

cal signal. Analysis results of these two approaches,

static on the one hand and dynamic on the other hand,

are reviewed hereinafter.

2.2 Electrocortical Potentials related to

Lower Limb Activation in Static

Condition

The cortical activity associated to bilateral anti-phase

and in-phase rhythmic foot movements produced by

subjects sitting on a chair was investigated in (Raeth-

jen et al., 2008). In this study, the authors found sig-

niﬁcant corticomuscular coherence between EEG sig-

nals and the anterior tibial muscles, at the stepping

frequencies in the central midline region, extending

further to the frontal mesial area. During isometric

cocontractionof the calf muscles, coherence appeared

between 15 and 30 Hz, concentrated on the central

midline area. This is the ﬁrst study demonstrating that

there exists a representation of rhythmic foot motor

patterns in the cortex, transmitted to the muscles and

fed back to the cortex with delays compatible with

fast corticospinal transmission, which may be impor-

tant for gait control.

Assisted lower limb movements have also been

investigated using electroencephalography (Wieser

et al., 2010). In this study, subjects performed stan-

dardized, assisted stepping movements (i.e. mim-

icking walk) in an upright position, while being se-

cured to a tilt table. Electrocortical sources associ-

ated to the movement-related potential were local-

ized in the primary motor cortex, the premotor cor-

tex, the supplementary motor cortex, the cingulate

cortex, the primary somatosensory cortex and the so-

matosensory association cortex (i.e. in accordance

with the ﬁndings of functional brain imagery). The

authors demonstrated that a clear succession of activa-

tions and deactivations was present in the movement-

related potential, in direct relationship with speciﬁc

phases of the gait-like leg movements. In particu-

lar, it was shown that cortical activity was the greatest

during transition between ﬂexion and extension of the

legs and vice versa.

Given the obvious possibility to detect electrocor-

tical potentials related to lower limb activation, two

studies were undertaken in order to develop brain-

computer interfaces (BCI) for motor augmentation.

In (Gwin and Ferris, 2011), it was shown that knee

contractions could be distinguished from ankle con-

tractions (subjects performed these exercises sitting

on a bench) using an independent analysis mixture

model applied on high-density EEG, without prior

knowledge of the exercise. An inverse modeling ap-

proach indicated the presence of electrocortical cur-

rent dipoles signiﬁcantly different for the knee and

ankle exercises. This ﬁnding is of course very promis-

ing for new applications in neurorehabilitation of gait

and control of robotic lower-limb exoskeletons.

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

310

In (Do et al., 2011), a non-invasive EEG-based

BCI governing a functional electrical stimulation

(FES) system for ankle movement is presented. In this

application, healthy subjects perform repetitive foot

dorsiﬂexions. EEG patterns underlying this action are

detected in real time, and this information is subse-

quently used to trigger the FES of the tibialis anterior

of the contralateral foot so as to achieve its dorsiﬂex-

ion. In fact, the trigger (or non-trigger) information

is given by a linear Bayesian classiﬁer trained using a

vector of spatio-spectral features which optimally dis-

criminate the idling and dorsiﬂexion states. The au-

thors state that analysis of subject-speciﬁc prediction

models demonstrated that the EEG power changes in

the µ, β and low γ bands observed over mid-central ar-

eas (i.e. electrode Cz) were the most informative fea-

tures for classiﬁcation. This likely corresponds to ac-

tivity within the primary motor cortex’s foot represen-

tation area and/or supplementary motor area (which is

not surprising from a brain anatomy standpoint) and

is in perfect agreement with prior studies (Neuper and

Pfurtscheller, 1996), (Solis-Escalante et al., 2008).

2.3 Electrocortical Potentials related to

Walk

To our knowledge, two very recent studies addressed

the dynamic approach of the problem (i.e. where sub-

jects are really walking). The ﬁrst analysis of EEG

during walk on treadmill was published by (Gwin

et al., 2011). By using a method based on independent

component analysis (ICA) combined with an inverse

modeling approach, the authors claimed they could

discriminate electrocortical sources, muscle sources

and other artifacts from the raw EEG signals. They

found that cortical activity in the anterior cingulate,

posterior parietal and sensorimotor cortex exhibited

signiﬁcant and smooth intra-stride changes in spectral

power. More precisely, alpha and beta band spectral

powers increased in or near the left/right sensorimo-

tor and dorsal anterior cingulate cortex at the end of

each stance phase (i.e. as the leading foot was con-

tacting the ground and the trailing foot was pushing

off). According to this study, power increases in the

left/right sensorimotor cortex were more important

for contralateral limb push-off (ipsilateral heel-strike)

than for ipsilateral limb push-off (contralateral heel-

strike). Finally, the authors reported evidence of intra-

stride high-gamma spectral power changes in anterior

cingulate, posterior parietal and sensorimotor cortex.

In parallel, (Presacco et al., 2011) showed for

the ﬁrst time that the kinematics of the ankle, knee

and hip joints during human treadmill walking can

be inferred from EEG signals. Successful decod-

ing of these signals was done basically by ﬁltering

them (0.1 – 2 Hz) and passing them through a lin-

ear autoregressive model. According to this study,

gait trajectories were inferred with accuracies com-

parable to those from neural decoders based on mul-

tiple single-unit activity recorded in non-human pri-

mates (Fitzsimmons et al., 2009). The results of

this study indicate a high involvement of a fronto-

posterior cortical network in the control of walking

and suggest that EEG signals can be used to study

in real time the cortical dynamics of walking and to

develop brain-machine interfaces aimed at restoring

human gait function.

3 DISCUSSION

3.1 About the Spatio-frequential

Characteristics of the Detected

Potentials

The results produced by the different analyses pre-

sented in previous section are in some way contra-

dictory. Table 1 summarizes the brain areas activated

during walk (or gait-like exercises) as well as the fre-

quency bands of interest, as reported by the different

authors.

From the spatial point of view, all the studies

found activations of the brain globally compatible

with the primary motor cortex’s foot representation

area and/or supplementary motor area, except one.

Surprisingly, (Presacco et al., 2011) report the acti-

vation of a complex, distributed and sparse cortical

network, in which scalp areas over anterior, right lat-

eral and right anterior-occipital scalp areas seem to

equally contribute (at least to their decoding of the

kinematics of the right leg, for subjects walking on a

treadmill).

From the frequential point of view, spectral power

variations were generally found from alpha to gamma

bands but, astonishingly, a successful neural decod-

ing of treadmill walk was realized by (Presacco et al.,

2011) using EEG signals band-pass ﬁltered between

0.1 and 2 Hz. This is particularly surprising, because

it was shown in another study, conducted to assess

EEG signal quality in motion environments (Kerick

et al., 2009), that EEG spectra in the walking (or

jogging) condition exhibit frequency peaks consistent

with the fundamental stride frequency as well as its

harmonics. The authors also state that motion arti-

facts affect signal integrity most prominently at low

frequencies (i.e. the delta band) during steady walk.

In their analysis protocol, (Presacco et al., 2011) do

EEG AND HUMAN LOCOMOTION - Descending Commands and Sensory Feedback should be Disentangled from

Artifacts Thanks to New Experimental Protocols Position Paper

311

Table 1: EEG studies of human locomotion: a schematic view of recent results obtained with static and dynamic experimental

protocols. M1 is the primary motor cortex, PMC is the premotor cortex, SMA is the supplementary motor cortex, CC is the

cingulate cortex, S1 is the primary somatosensory cortex and SA is the somatosensory association cortex.

Publication Aim of the study Approach Activated brain areas Frequency bands of interest

Raethjen et al., 2008 Rhythmic foot move-

ments

Static Central midline region

+ frontal mesial area

Stepping frequency + β

band (15 – 30 Hz)

Wieser et al., 2010 Assisted lower-limb

movements

Static M1, PMC, SMA, CC,

S1, SA

No frequency analysis. Ac-

tivations are directly related

to speciﬁc phases of the

gait-like movements

Do et al., 2011 BCI dedicated to a FES

system for ankle move-

ment

Static Mid-central areas (elec-

trode Cz)

µ, β and low-γ bands

Gwin et al., 2011 EEG activity during

treadmill walking

Dynamic Anterior cingulate, pos-

terior parietal and sen-

sorimotor cortex

α and β bands + clear evi-

dence of high-γ intra-stride

spectral power changes

Presacco et al., 2011 Neural decoding of

treadmill walking from

EEG signals

Dynamic Involvement of a

broad fronto-posterior

cortical network

Delta band (0.1 – 2 Hz)

not mention any pre-processing method aiming at ei-

ther correcting or discriminating these motion arti-

facts from the real cortical signals. The only way

for them to make the choice of this frequency band

legitimate is the fact that good results are obtained

and, moreover, other studies exploited the same por-

tion of the EEG spectrum to decode upper limb move-

ments. We strongly emphasize the fact that, in the

latter studies, no motion artifacts due to gait are pro-

duced. Consequently, this might suggest that the de-

coding of kinematics of walk – periodical movement

– on the basis of the EEG signals is done by (Pre-

sacco et al., 2011) with a linear autoregressive model

exploiting the periodical motion artifacts present in

the EEG recordings. This option is furthermore sup-

ported by the fact that no spectral information is given

under 3 Hz in the study of (Gwin et al., 2010).

3.2 About the Origin of the Detected

Signals

Among all the works described in section 2, only

(Raethjen et al., 2008) try to determine the origin

of the information ﬂux contained in the studied sig-

nals (descending commands from the brain or sen-

sory feedback sent to the brain). This is done by

computing time delays between EEG time series and

electromyographicactivity of the involved lower limb

muscles by means of the “maximising coherence

method” (Govindan et al., 2005). Actually, other

studies presented in section 2 do not consider this as-

pect and give no indication on the direction of the

brain-muscle interaction (i.e. if it is up-going or

down-going). It is therefore unknown, for instance,

if the intra-strides spectral power variations found by

(Gwin et al., 2010) are due to voluntary movements

or sensory feedback (or a combination of both). The

same question arises concerning the EEG decoding

presented by (Presacco et al., 2011). Resolving this

ambiguity is particularly crucial, though, for the de-

velopment of gait rehabilitation systems. Indeed, if

the information detected in the EEG signals is purely

due to a sensory feedback of the gait-related move-

ments, it would be unusable to drive any device, given

that no valid prediction of a movement can be done

exploiting sensory information resulting from it.

Most importantly, studying EEG signals in tread-

mill walking also requires the need to exclude gait-

related artifacts. Only one study tackles this issue

(Gwin et al., 2011). Using an ICA analysis cou-

pled with an inverse solution approach, these authors

claim that they could disentangle muscular contribu-

tions and other artifacts from real cortical signals.

However, in a previous study, the very same authors

(Gwin et al., 2010) clearly stated that:

“Unlike more spatially stationary artifacts in EEG

signals arising from eye movements, scalp muscles,

fMRI gradients, etc., which may be resolved by ICA

decomposition into a subspace of one or more in-

dependent components, we found that gait-related

movement artifact remained in many if not most of

the independent components. This prevented us from

removing only a small subset of components captur-

ing the movement artifacts.”

For this reason, they considered the removal of

motion artifacts from EEG during walking and run-

ning on treadmill using an artifact template subtrac-

tion method. Such method allowed to enhance the de-

tection of P300 potentials in ambulatory conditions.

Nevertheless, the study of cerebral processes involved

in human locomotion is not possible using a subtrac-

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

312

Figure 1: Gait event-related spectral perturbation (ERSP)

plots showing average changes in spectral power during the

stride cycle relative to the full gait cycle baseline for differ-

ent brain areas (Figure from (Gwin et al., 2011)).

tion method, as it would undoubtedly remove inter-

esting signal from the EEG recordings. For this rea-

son, the authors used only the ICA approach to clean

the EEG signals (Gwin et al., 2011). In this study,

the issue of motion artifacts was completely eluded

and no mention was made of any appropriate treat-

ment to reject them. Thus, it can be doubted that the

time-frequency analysis plots shown in that paper do

not contain any motion artifact contribution (see Fig-

ure 1). Figure 2 shows, for instance, a time-frequency

analysis of the signal of an accelerometer placed on

the head of a subject walking at 1.67 m/s on a tread-

mill. Periodic power spectral changes over large fre-

quency bands can be observed, in a similar way to the

results obtained after ICA (see Figure 1) by (Gwin

et al., 2011).

4 FUTURE WORK

Several contradictory results coming from different

recent EEG studies of human locomotion have been

discussed. We have shown that the discrepancies con-

cern the spatio-frequential characteristics of the de-

tected potentials and that the presence of motion arti-

facts is not to be excluded in certain studies.

In this context, it seems necessary to deﬁne sev-

eral experimental protocols in order to disentangle the

different signals (are they associated to descending or

ascending pathways, are they artifacts, ...).

First, we propose to characterize the descending

brain commands which are involved in human walk

control in a static approach (inspired by (Raethjen

Figure 2: Representative ERSP plot obtained with the out-

put signal of an accelerometer placed on the head (thus

undergoing the same shocks as the EEG electrodes) of a

subject walking at 1.67 m/s on a treadmill. The horizon-

tal axis is time (and not a percentage of gait cycle like in

Figure 1) as no time-warping analysis was done. Reference

time (t=0) corresponds to the left heel strike (LON) instant.

The same alternating spectral power changes are also ob-

served at other speeds.

et al., 2008)), in order to ensure absence of EEG me-

chanical artifacts. To this aim, the EEG signals of sub-

jects sitting on a chair will be recorded. The subjects

will then be asked to produce voluntary rhythmic foot

movements, staying at the same tempo. The feet will

not be in contact with the ground, to ensure a mini-

mal sensorimotor feedback. Several tempos will be

produced. We will also record EEG when the sub-

ject is sitting and not moving the feet, to deﬁne a

baseline, necessary when using brain imagery tools

like LORETA (Pascual-Marqui, 2002). To assess the

presence (or absence) of mechanical artifacts, an ac-

celerometer will be placed on his neck. A complete

characterization of these data will then be realized,

by analysing the event-related spectral perturbations

(ERSP) combined with a time-warping transforma-

tion (Gwin et al., 2011), and by computing cortico-

muscular coherence and in particular delays between

EEG and EMG time series (to assess the information

ﬂow direction).

Then, we will characterize EEG signals caused by

somatosensory information coming from the feet of

the subject when sitting (again, to prevent any me-

chanical artifacts). More precisely, the same experi-

ment as above will be realized, with the feet in contact

with ground, this time. By comparing the two states

(contact/no contact), it will be possible to emphasize

the contribution of sensory feedback. Alternatively,

we intend to use special tactors to stimulate the feet,

mimicking the sensation of walk and will study the

properties of the EEG signals that are phase-locked

with this stimulation.

Additionally, in order to characterize the motion

artifacts contribution in an independent way, we pro-

EEG AND HUMAN LOCOMOTION - Descending Commands and Sensory Feedback should be Disentangled from

Artifacts Thanks to New Experimental Protocols Position Paper

313

pose to use an EEG test setup analog to the one de-

scribed in (Nonclercq and Mathys, 2010), compris-

ing a generator that produces cerebral-like waves, a

dummy head, an electrode/gel/skin interface model,

electrodes, and leads. Placing this setup in the ref-

erence frame of a subject walking on a treadmill

would produce realistic gait-related motion artifacts

(it would indeed undergo the same shocks as the EEG

electrodes) and should give valuable information to

subsequently reject them.

Finally, if we correctly reject motion artifacts, pro-

vided we know the signals due to descending com-

mands (voluntary rhythmic movements) and those

due to tactile stimulation (tactors, mimicking the sen-

sation of walk), we should be able to disentangle the

contribution of posture and balance control when the

subject is standing and walking.

ACKNOWLEDGEMENTS

This work was funded by the FEDER support (BIO-

FACT). M. Duvinage is a FNRS (Fonds National de

la Recherche Scientiﬁque) Research Fellow. This

paper presents research results of the Belgian Net-

work DYSCO (Dynamical Systems, Control, and Op-

timization), funded by the Interuniversity Attraction

Poles Programme, initiated by the Belgian State, Sci-

ence Policy Ofﬁce. The scientiﬁc responsibility rests

with its author(s).

REFERENCES

Castermans, T., Duvinage, M., and Dutoit, T. (2011). Op-

timizing the performances of a P300-based Brain-

Computer Interface in ambulatory conditions. IEEE

Journal on Emerging and Selected Topics in Circuits

and Systems (JETCAS), Special Issue on ”Brain Ma-

chine Interfaces” (in Press).

Do, A.-H., Wang, P.-T., King, C.-E., Abiri, A., and Nenadic,

Z. (2011). Brain computer interface controlled func-

tional electrical stimulation system for ankle move-

ment. Journal of NeuroEngineering and Rehabilita-

tion, 8(49).

Fitzsimmons, N. A., Lebedev, M. A., Peikon, I. D., and

Nicolelis, M. A. L. (2009). Extracting kinematic pa-

rameters for monkey bipedal walking from cortical

neuronal ensemble activity. Frontiers in integrative

neuroscience, 3(March):19.

Govindan, R., Raethjen, J., Kopper, F., Claussen, J., and

Deuschl, G. (2005). Estimation of time delay by co-

herence analysis. Physica A: Statistical Mechanics

and its Applications, 350(2-4):277 – 295.

Gwin, J. and Ferris, D. (2011). High-density EEG and in-

dependent component analysis mixture models distin-

guish knee contractions from ankle contractions. In

Proc. of the 33rd Annual International Conference of

the IEEE EMBS, Boston, Massachussets, USA, pages

4195 – 4198.

Gwin, J., Gramann, K., Makeig, S., and Ferris, D. (2010).

Removal of movement artifact from high-density EEG

recorded during walking and running. Journal of Neu-

rophysiology, 103(6):3526–3534.

Gwin, J., Gramann, K., Makeig, S., and Ferris, D. (2011).

Electrocortical activity is coupled to gait cycle phase

during treadmill walking. NeuroImage, 54(2):1289 –

1296.

Hanakawa, T. (2006). Neuroimaging of standing and walk-

ing: Special emphasis on parkinsonian gait. Parkin-

sonism and Related Disorders, 12(Supplement 2):S70

– S75. Proceedings of the 1st International Sympo-

sium on Dopaminergic and Nondopaminergic Mecha-

nisms in Parkinson’s Disease (ISDNMPD).

Harada, T., Miyai, I., Suzuki, M., and Kubota, K. (2009).

Gait capacity affects cortical activation patterns re-

lated to speed control in the elderly. Experimental

Brain Research, 193(3):445–454.

Kerick, S. E., Oie, K. S., and McDowell, K. (2009). Assess-

ment of EEG signal quality in motion environments.

Army Research Laboratory Report.

Kiehn, O. (2006). Locomotor circuits in the mammalian

spinal cord. Ann. Rev. of Neuroscience, 29:279 – 306.

Neuper, C. and Pfurtscheller, G. (1996). Post-movement

synchronization of beta rhythms in the EEG over

the cortical foot area in man. Neuroscience Letters,

216(1):17 – 20.

Nonclercq, A. and Mathys, P. (2010). Quantiﬁcation of the

movement artifact reduction in EEG recording. IEEE

trans. on Biom. Eng., 57(11):2746 – 2752.

Nunez, P., Srinivasan, R., Westdorp, A., Wijesinghe, R.,

Tucker, D., Silberstein, R., and Cadusch, P. (1997).

EEG coherency. Electroencephalography and Clini-

cal Neurophysiology, 103(5):499–515.

Pascual-Marqui, R. (2002). Standardized low-resolution

brain electromagnetic tomography (sloreta): techni-

cal details. Methods Find. Exp. Clin. Pharmacol.,

24(D):5–12.

Presacco, A., Goodman, R., Forrester, L., and Contreras-

Vidal, J. (2011). Neural decoding of treadmill walking

from non-invasive, electroencephalographic (EEG)

signals. Journal of Neurophysiology. in press.

Raethjen, J., Govindan, R., Binder, S., Zeuner, K. E.,

Deuschl, G., and Stolze, H. (2008). Cortical represen-

tation of rhythmic foot movements. Brain Research,

1236:79 – 84.

Solis-Escalante, T., M¨uller-Putz, G., and Pfurtscheller, G.

(2008). Overt foot movement detection in one single

laplacian EEG derivation. Journal of Neuroscience

Methods, 175(1):148 – 153.

Suzuki, M., Miyai, I., Ono, T., and Kubota, K. (2008). Ac-

tivities in the frontal cortex and gait performance are

modulated by preparation. An fNIRS study. Neuroim-

age, 39(2):600 – 607.

Wieser, M., Haefeli, J., B¨utler, L., J¨ancke, L., Riener, R.,

and Koeneke, S. (2010). Temporal and spatial pat-

terns of cortical activation during assisted lower limb

movement. Exp. Brain Res., 203:181–191.

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

314