Phase-locked and Non-phase-locked Event-related Oscillations and Channel Power Spectra Analysis during Motor Imagery with Speed Parameters for

BCRI

Yunfa Fu1,2, Baolei Xu1,2, Lili Pei1,2, Hongyi Li1 1. State Key Laboratory of Robotics

Shenyang Institute of Automation, Chinese Academy of Sciences Shenyang, China

2. Graduate School of the Chinese Academy of Sciences Beijing, China

fuyunfa@sia.cn

Abstract—Phase-locked and non-phase-locked event-related oscillations and channel spectra during motor imagery with speed parameters (fast 4Hz and slow 1Hz) were investigated in the paper. EEG signals related to imagination of six tasks that involved three limbs (left and right index fingers and right toes) and two speeds were first separated from original EEG signals mixed with noises by ICA. Then, event-related EEG trials were superposed and averaged to get time domain average EEG that was decomposed by wavelet transform to investigate phase-locked event-related oscillations. Event-related EEG trials were also decomposed by wavelet transform respectively and then superposed and averaged to investigate non-phase-locked event-related oscillations. In addition, power spectra for selected channel were calculated by FFT. By these methods, the latency of 50~536ms and the low frequency band of 2.7~8Hz for an increase in power early after the event, latencies and frequency bands for significant different changes and a different degree decrease in power at high frequency band between fast and slow motor imagery, and ERD / ERS in a broad frequency band for selected channels were found. Simultaneously extracting phase-locked and non-phase-locked event-related oscillations in broadband range may be very important for acquiring complete information about event-related EEG during motor imagery with speed parameters. The research results may be used in feature extraction for brain-controlled robots interface (BCRI). In the based on brain cognitive robots control, this study is also expected to provide some degree continuous and fine control strategies for BCR.

Keywords: Motor imagination with speed parameters; Phase-locked event-related oscillations; Non-phase-locked event-related oscillations; ERD/ERS; Brain-controlled robots interface (BCRI); brain-controlled robots (BCR)

I. INTRODUCTION Brain-controlled robots interface (BCRI) based on brain

cognition without muscle activity involvement is an interface technique between human brain and robots that not only can help severely motor-disabled persons to control their environment to improve the quality of their lives but also can assist healthy persons to accomplish certain tasks under

specific circumstances [1]-[10]. The final objective of BCRI is to execute a certain extent continuous and fine control of robots as well as simple switch variables or directions control [11], [12]. It is necessary for this goal to not only identify the body type of motor imagery but also decode parameters variation in motor imagery with speed and force by BCRI.

There are more studies in identifying limb types of motor imagery than decoding motor imagery with speed and force parameters [13]-[15]. In addition, decoding subject’s mental intentions by measuring electrical activities of neurons or neuron ensemble or neural network based on invasive methods with potential injuries, which have very high spatial and temporal resolution, had been proved feasible and made progress [16],[17]. Decoding parameters variation of motor imagery by measuring EEG resulted from the activities of a large number of neurons based on non-invasive methods with electrodes placed on human scalp, which have low spatial resolution, is also one of the most challenging research works, and a few studies have showed the possibility [13]-[15]. However, how to improve the spatial resolution of EEG by inventing a new non-invasive measuring method and how to overcome the bottleneck of low spatial resolution of EEG by designing advanced signal processing technologies are still technical problems to be solved. Another solution is expected to propose appropriate methods to fuse high level task commands provided by BCRI and intelligent robot’s capability to achieve continuous and fine control of robots [18].

From research methods for event-related oscillations, average ERPs obtained by traditional time domain average superposition method that can't extract complete brainwave oscillation information only contains phase-locked evoked rhythm excluding non-phase-locked induced rhythm [19],[20]. Therefore, in this study, we based on a new experimental paradigms used in BCRI research proposed by us and EEG recorded from electrodes placed on the scalp to investigate simultaneously phase-locked event-related oscillations and non-phase-locked event-related oscillations during motor imagery at fast and slow.

The work was supported by the National Natural Science Foundation of China (Grant No. 60705021) and the research project of State Key Laboratory of Robotics of Shenyang Institute of Automation (SIA), Chinese Academy of Science (CAS) (Grant No. 08A120C101).

978-1-4244-5089-3/11/$26.00 ©2011 IEEE

II. METHODS

A. EEG Data Collecting EEG signal acquisition was based on a special design

paradigm that was oriented to continuous and fine control of robots. In the new paradigm, motor imagery was planned to involve different limbs and speeds and make subjects easily execute them. In the study, EEG data were acquired from 4 healthy subjects who were instructed to imagine tapping table (or left mouse button) with their left and right index fingers and tapping floor with their right toes at two speeds (fast 4Hz and slow 1Hz). The experimenter guided them to execute actual six tasks including Left Forefinger Fast (LFF), Left Forefinger Slow (LFS), Right Forefinger Fast (RFF), Right Forefinger Slow (RFS) Right Toes Fast (RTF), and Right Toes Slow (RTS) so as to obtain experience before signal acquisition. During the experiment, subjects performed motor imagination at a first-person perspective with recalling and feeling the kinesthetic experience of movement (first imaging). The six tasks were presented randomly to the subjects. During motor imagination, no visual feedback which indicated effect of task was provided to subjects who were asked to keep relaxation and avoid muscle activation, blink, slow eye movement, and facial muscle tension. Each subject took part in three sessions of 4 runs of 60 trials. Timing of a single trial is referenced in [21], [22].

We used a 64-channel digital DC EEG amplifier (Neuro Scan Labs, synAmps 2 ) and signal band pass 0.05-100 Hz and a 50 Hz notch filter to acquire EEG signals (sampling frequency at 500Hz with a 24-bit A/D converter). Ag-AgCl electrodes (extended 10-20 system) and GND at forehead and REF at vertex were allocated in the experiment.

B. EEG Data Preprocessing The objectives of preprocessing original EEG data lie in

getting relatively clean task related EEG epoch signals for further analyses. The procedure of EEG data preprocessing is shown in Fig. 1.The analysis tool in this research is EEGLAB [23].

C. Phase-locked Event-related Oscillations Analysis Channel selection based on the study involved the brain

areas is very important because of the identification accuracy and practicality. The reasons we choose to channels for further analysis as follows.

On the one hand, almost all areas of the neocortex such as the prefrontal cortex and the primary somatosensory cortex (S1) as well as the cerebellum and subcortical motor nuclei outside the cortex play a role in the control of voluntary movements, but one of the brain areas dedicated to controlling these voluntary movements is the motor cortex [24].The motor cortex can be divided into two groups with four main parts: (1) the primary motor cortex (or M1), which is responsible for generating the neural impulses controlling execution of movement and (2) the secondary motor cortices, including the posterior parietal cortex that is responsible for transforming visual information into motor command and the premotor cortex that is responsible for motor guidance of movement and

Figure 1. The procedure of EEG data preprocessing

control of proximal and trunk muscles of the body and the supplementary motor area (SMA) that is responsible for planning and coordination of complex movements such as those requiring two hands. The main cortices involved in the control of voluntary movements are shown in Fig. 2 [25]. Planning for any given movement is done mainly in the forward portion of the frontal lobe which receives information about the individual's current position from several other parts. Then it issues its commands to Area 6 (supplementary motor and premotor cortex area) which decides which set of muscles to contract to achieve the required movement, and then issues the corresponding orders to the primary motor cortex (Area 4) which in turn activates specific muscles or groups of muscles via the motor neurons in the spinal cord [24].

Figure 2. The main cortices involved in the control of voluntary movements

On the other hand, not only actual voluntary motor is controlled and executed by motor cortex, some famous studies have shown that motor imagery can modify the neuronal activity in the primary sensorimotor areas in a very similar

Down-sampling original EEG signals at 250Hz

Converting reference electrode from vertex to bilateral mastoid reference (M1, M2)

Visually inspecting obvious artifacts excluded from further analyses

Extracting epoch data from -2 to 4 s by EEGLAB

Eliminating EOG and EMG artifact by ICA with EEGLAB

Excluding bad epochs for further analyses

way as observable with a real executed movement [26]. Features such as band power or adaptive auotoregressive parameters are either extracted in bipolar EEG recordings overlaying sensorimotor areas or from an array of electrodes located over central and neighboring areas [26].

Based on the above comprehensive consideration, the channels C3, CZ, and C4 which overlap motor cortex area were selected to explore motor imagery with parameters as shown in Fig. 3.

C3 CZ C4

Figure 3. Channel locations for phase-locked and non-locked phase event-related oscillations analysis

The traditional average ERP technology was applied to get phase-locked event-related oscillations which in a certain extent can reflects the dynamic process of brain cognition in the study. In addition to this, the basic and key properties of non-stationary EEG signal are time domain local information. Therefore, time-frequency analysis can be a suitable tool for EEG [20].The power spectral of average event-related EEG at frequency f and time t is estimated by short-time Fourier transform:

τττ

τττ τπ

dgf

detgftfF

Rtf

R

fi

)()(

)()(),(

*,

2

∫

∫=

−= −

(1)

where τπττ fitf etgg 2*

, )()( −−= as base function, window

function )( tg −τ as time limit and shift, τπfie 2− as frequency limit. In the study, ),( tfF is calculated by EEGLAB with a sinusoidal wavelet (short-time DFT) transformation in which the number of cycles is increased slowly with frequency to obtain better frequency resolution at higher frequencies than a conventional wavelet approach that uses constant cycle length [23].

The procedure for phase-locked event-related oscillation analysis using the EEGLAB V9_0_0_2b toolbox (at http: //sccn.ucsd.edu/eeglab/) under MATLAB v7.6.0 (Mathworks, Natick, MA) as shown in Fig. 4.

D. Non-phase-locked Event-related Oscillations Analysis In order to extract non-phase-locked event-related

Figure 4. The procedure for phase-locked event-related oscillations analysis

oscillations, the time-frequency power ( , )kF f t of a single trial k was calculated based on (1), and then event-related spectral perturbation (ERSP) as follows [23]:

∑=

=n

kk tfF

ntfERSP

1

2),(1),( (2)

where n is the total of trials. Significance of deviations from baseline power is assessed using a bootstrap method [23].0.01 is for the bootstrap significance level in the paper. Spectral power starts from -2000ms related to stimulus, baseline limits: -2000~-1000ms, epoch time limits:-2000~4000 ms, and frequency limits: <50Hz.The procedure for non-phase-locked event-related oscillations analysis as shown in Fig. 5.

Figure 5. The procedure for non-phase-locked event-related oscillations

analysis

E. Channels Power Spectral Analysis ERD/ERS patterns reflecting sensorimotor activation and

deactivation are very important for BCRI based on motor imagery [27]-[29]. In order to explore power spectral distribution on the scalp at specified frequency 12, 13, 22, 26, 29, 30, 34, 40Hz, the spectra and map for chosen frequency and chosen channel at C3, Cz, C4 were calculated and presented by the EEGLAB V9_0_0_2b toolbox.

III. RESULTS We used methods in sectionⅡ to analyze EEG data from 4

Preprocessing EEG data

Superposing and averaging event-related EGG trials

Calculating the time-frequency power of average event-related EEG

Event-related EEG trials

Preprocessing EEG data

Calculating the time-frequency power of single event-related EEG trial

Superposing and averaging the time-frequency power of event-related EEG trials

Event-related EEG trials

subjects and found similar results among subjects. Therefore, the representative results from subject1 are as follows.

A. The Results of Phase-locked Event-related Oscillations Analysis for Fast and Slow Motor Imagination.

Fig. 6 to 11 present the results of phase-locked event-related oscillation analysis for subject1 (subj1) during fast and slow motor imagery at the first view. Following TABLE I from the above figures shows latencies and low frequency bands for an increase in power early after the event under motor imagery modes. From TABLE I, common latency is about 50 ~ 536ms and frequency band is about 2.7-8Hz.This

ERSP(dB)

-20

0

20

-1000 -500 0 500 1000 1500 2000 2500 3000-40

20

Time (ms)

dB10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC phase Figure 6. Phase-locked event-related oscillations analysis for subj1-first

imaging-RFF-C3

ERSP(dB)

-20

0

20

-1000 -500 0 500 1000 1500 2000 2500 3000-40

20

Time (ms)

dB

10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC phase Figure 7. Phase-locked event-related oscillations analysis for subj1-first

imaging-RFS-C3

ERSP(dB)

-20

0

20

-1000 -500 0 500 1000 1500 2000 2500 3000

-40

20

Time (ms)

dB

10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC phase Figure 8. Phase-locked event-related oscillations analysis for subj1-first

imaging-LFF-C4

ERSP(dB)

-20

0

20

-1000 -500 0 500 1000 1500 2000 2500 3000-60

20

Time (ms)

dB

10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC phase Figure 9. Phase-locked event-related oscillations analysis for subj1-first

imaging-LFS-C4

ERSP(dB)

-20

0

20

-1000 -500 0 500 1000 1500 2000 2500 3000-60

40

Time (ms)

dB

10

20

30

40

50

0 20

Fre

quen

cy (

Hz)

dB

ITC phase Figure 10. Phase-locked event-related oscillations analysis for subj1-first

imaging-RTF-Cz

ERSP(dB)

-20

0

20

-1000 -500 0 500 1000 1500 2000 2500 3000-60

20

Time (ms)

dB

10

20

30

40

50

0 20

Fre

quen

cy (

Hz)

dB

ITC phase Figure 11. Phase-locked event-related oscillations analysis for subj1-first

imaging-RTS-Cz

TABLE I. THE LATENCIES AND LOW FREQUENCY BAND FOR AN INCREASE IN POWER EARLY AFTER THE EVENT UNDER MOTOR IMAGERY

MODES

Imagery modeand channels

Latency (ms )

Low frequencyBand (Hz)

Brainactivity

RFF-C3 0~536 2.7~12

An increase in power

early after the event

RFS-C3 0~536 2.7~7.5

LFF-C4 50~536 2.7~9

LFS-C4 50 ~600 2.7~9.5

RTF-Cz 50~750 2.7~13

RTS-Cz 50~600 2.7~8.5

may be a very valuable result for BCRI based on fast and slow motor imagery.

In addition to the above important common early phase-locked event-elated oscillation, the latencies and frequency bands for significant different changes in power between fast and slow motor imagery from the above figures are shown in TABLE II These may be also very important results for further identification of a single trial

B. The Results of Non-phase-locked Event-related Oscillations Analysis for Fast and Slow Motor Imagination

Fig. 12 to 17 shows the results of non-phase-locked event-elated oscillation analysis for subject1 during fast and slow motor imagery at the first view. Surprisingly, we still can get phase-locked event-related components whose patterns (frequency band and latency) are very similar to the results by phase-locked event-related oscillation analysis whereas only their intensity becomes weaker than the later. Therefore, this may further show that phase-locked event-related component is a relatively stable oscillatory and can be used as classification feature for motor imagery at fast and slow. In

addition, Fig. 12 to 17 also shows an important result that a different degree decrease in power occurs at high frequency band.

TABLE II. THE LATENCIES AND FREQUENCY BANDS FOR SIGNIFICANT DIFFERENT CHANGES IN POWER BETWEEN FAST AND SLOW MOTOR IMAGERY

Imagery mode & channel

Latency (ms )

Frequency band (Hz)

Power changesFor brain activity

RFF-C3/ RFS-C3 -1000~0 6.6~20 Incre/ Decre

1400~2900 2.7~13 Incre/ Decre

LFF-C4/ LFS-C4

-1000~4000 above 25 Decre/ incre

-1000~-200 12~14.4 incre /Decre

730~4000 2.7~12.5 Decre/ incre

RTF-Cz/ RTS-Cza

750~4000 40~44.4 Incre/ decre

1000~4000 26.8~32.5 Decre incre

-950~0 7.5~13.5

Incre/ decre

800~4000 Incre decre

a Incre/ decre for RTF-Cz/ RTS-Cz denote an increase in power for RTF-Cz and a decrease in power for RTS-Cz. Similar to the other.

ERSP(dB

-5

0

5

-1000 -500 0 500 1000 1500 2000 2500 3000-5

5

Time (ms)

dB

10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC Figure 12. Non-phase-locked event-related oscillations analysis for subj1-first

imaging-RFF-C3

ERSP(dB

-5

0

5

-1000 -500 0 500 1000 1500 2000 2500 3000-5

5

Time (ms)

dB

10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC Figure 13. Non-phase-locked event-related oscillations analysis for subj1-first

imaging-RFS-C3

ERSP(dB

-5

0

5

-1000 -500 0 500 1000 1500 2000 2500 3000-10

5

Time (ms)

dB

10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC

Figure 14. Non-phase-locked event-related oscillations analysis for subj1-first imaging-LFF-C4

ERSP(dB

-5

0

5

-1000 -500 0 500 1000 1500 2000 2500 3000-5

5

Time (ms)

dB

10

20

30

40

50

20 40

Fre

quen

cy (

Hz)

dB

ITC Figure 15. Non-phase-locked event-related oscillations analysis for subj1-first

imaging-LFS-C4

ERSP(dB

-5

0

5

-1000 -500 0 500 1000 1500 2000 2500 3000-5

5

Time (ms)

dB

10

20

30

40

50

2040

Fre

quen

cy (

Hz)

dB

ITC Figure 16. Non-phase-locked event-related oscillations analysis for subj1-first

imaging-RTF-Cz

ERSP(dB

-4

-2

0

2

4

-1000 -500 0 500 1000 1500 2000 2500 3000-5

5

Time (ms)

dB

10

20

30

40

50

2040

Fre

quen

cy (

Hz)

dB

ITC Figure 17. Non-phase-locked event-related oscillations analysis for subj1-first

imaging-RTS-Cz

C. The Results of Channel Power Spectra Analysis. Fig. 18 and 19 show a decrease in power (ERD) at 12.2

(12) and 13.2 (13) Hz at left brain with blue color during motor imagery at fast and slow for right index finger. But a greater increase in power at these two frequencies at right brain with red color at slow than at fast occurs. In addition, in the case of higher frequency 22, 25.9(26), 28.8(29), 29.8(30), 34.2(34),40Hz,an increase in power (ERS) at left brain for RFF but still ERD at right brain for RFS except for 40Hz.

Fig. 20 and 21 show a decrease in power (ERD) at 12.2 (12) and 13.2 (13) Hz at right brain with blue color during motor imagery at fast and slow for left index finger. A slight decrease in power at 22,25.9(26),28.8(29),29.8(30),34.2(34)

0 10 20 30 40 50 60

-20

-10

0

10

20

30

Frequency (Hz)

Pow

er 1

0*lo

g 10(μ

V2 /H

z)

12.2 13.2 22.0 25.9 28.8 29.8 34.2 40.0 Hz

Figure 18. Channel spectra and maps for subj1-first imaging-RFF

0 10 20 30 40 50 60

-20

-10

0

10

20

30

Frequency (Hz)

Pow

er 1

0*lo

g 10(μ

V2 /H

z)

12.2 13.2 22.0 25.9 28.8 29.8 34.2 40.0 Hz

Figure 19. Channel spectra and maps for subj1-first imaging-RFS

Hz at vertex area with light blue color and an increase in power at 40Hz at right brain for slow motor imagery whereas a decrease in power at 22, 25.9(26), 40Hz at right brain for fast motor imagery.

0 10 20 30 40 50 60

-20

-10

0

10

20

30

Frequency (Hz)

Pow

er 1

0*lo

g 10(μ

V2 /H

z)

12.2 13.2 22.0 25.9 28.8 29.8 34.2 40.0 Hz

Figure 20. Channel spectra and maps for subj1-first imaging-LFF

0 10 20 30 40 50 60

-20

-10

0

10

20

30

Frequency (Hz)

Pow

er 1

0*lo

g 10(μ

V2 /H

z)

12.2 13.2 22.0 25.9 28.8 29.8 34.2 40.0 Hz

Figure 21. Channel spectra and maps for subj1-first imaging-LFS

Fig. 22 and 23 show a decrease in power at 12.2 (12) and 13.2 (13) Hz at vertex area with blue color during motor imagery at fast and slow for right toes. An increase in power at 22, 40 Hz at left brain area during motor imagery at fast is not the same as motor imagery at slow.

0 10 20 30 40 50 60

-20

-10

0

10

20

30

Frequency (Hz)

Pow

er 1

0*lo

g 10(μ

V2 /H

z)

12.2 13.2 22.0 25.9 28.8 29.8 34.2 40.0 Hz

Figure 22. Channel spectra and maps for subj1-first imaging-RTF

0 10 20 30 40 50 60

-20

-10

0

10

20

30

Frequency (Hz)

Pow

er 1

0*lo

g 10(μ

V2 /H

z)

12.2 13.2 22.0 25.9 28.8 29.8 34.2 40.0 Hz

Figure 23. Channel spectra and maps for subj1-first imaging-RTS

IV. CONCLUSION In the study, based on a new experimental paradigm of fast

and slow motor imagery involved three limbs used in BCRI research, we selected only three electrode locations C3, Cz, C4 of motor cortex area closely related to movement imagination, and then the latency of 50 ~ 536ms and the low

frequency band of 2.7~8Hz for an increase in power early after the event were shown not only by phase-locked event-related analysis but also non-phase-locked event-related analysis. In addition, latencies and frequency bands for significant different changes in power between fast and slow motor imagery and a different degree decrease in power at high frequency band also were shown using these two kinds of methods. Also, we showed ERD and ERS in a broad frequency band for selected channels. This study will provide basic support for the realization of brain-controlled robots interface based on human brain cognitive and make continuous and fine control of robots by BCRI possible. The study also will provide methods and techniques support for the new research for the integration and fusion of human brain and robots. In the future BCRI research, we will control a humanoid robot in our laboratory by identification of single trial during fast and slow motor imagery.

ACKNOWLEDGMENT The authors are grateful to Dr. Lun Zhao, Yongcheng Li

Shujia Qin, Wei Ding, and Lei Miao for helpful discussions. Also the authors would like to thank Lijie Dang and Tongran Liu for assistance in acquiring the experiment data .

REFERENCES [1] Christian J Bell, Pradeep Shenoy, Rawichote Chalodhornand Rajesh P N

Rao. Control of a humanoid robot by a noninvasive brain–computer interface in humans. J. Neural Eng. 5 (2008) 214–220

[2] Muhammad Nabeel Anwar, Vittorio Sanguineti, Pietro Giovanni Morasso, Koji Ito. Motor imagery in robot-assistive rehabilitation: A study with healthy subjects 2009 IEEE 11th International Conference on Rehabilitation Robotics Kyoto International Conference Center, Japan, June 23-26, 2009. 337-342

[3] Inaki Iturrate,Javier M. Antelis, Andrea Kubler, and Javier Minguez. A Noninvasive Brain-Actuated Wheelchair Based on a P300 Neurophysiological Protocol and Automated Navigation. IEEE transactions on robotics, vol. 25, no. 3, june 2009 614-627

[4] Eduardo I′a˜nez, M. Clara Furi′o, Jos′e M. Azor′ın, Jos′e Alejandro Huizzi, and Eduardo Fern′andez. Brain-Robot Interface for Controlling a Remote Robot. ArmJ. Mira et al. (Eds.): IWINAC 2009, Part II, LNCS 5602, pp. 353–361.

[5] J. del R. Mill′an, F. Gal′an, D. Vanhooydonck, E. Lew, J. Philips and M. Nuttin. Asynchronous Non-Invasive Brain-Actuated Control of an Intelligent Wheelchair. 31st Annual International Conference of the IEEE EMBS Minneapolis, Minnesota, USA, September 2-6, 2009. 3361-3364

[6] Dennis J. McFarland and Jonathan R. Wolpaw. Brain-Computer Interface Operation of Robotic and Prosthetic Devices. Computer, October 2008,52-56. www.computer.org/join/grades.htm

[7] Tao Geng and John Q. Gan Motor Prediction in Brain-Computer Interfaces for Controlling Mobile Robots 30th Annual International IEEE EMBS Conference, Vancouver, British Columbia, Canada, August 20-24, 2008,634-637.

[8] Mufti Mahmud , David Hawellek, Aleksander Valjamae. A Brain-Machine Interface Based on EEG: Extracted Alpha Waves Applied to Mobile Robot. 2009 Advanced Technologies for Enhanced Quality of Life,28-31.

[9] Alexandre O. G. Barbosa, David R. Achanccaray, and Marco A. Meggiolaro .Activation of a Mobile Robot through a Brain Computer Interface. 2010 IEEE International Conference on Robotics and Automation.Anchorage Convention DistrictMay 3-8, 2010, Anchorage, Alaska, USA,4815-4825.

[10] José del R. Millán, Frédéric Renkens, Josep Mouriño, and Wulfram Gerstner .Noninvasive Brain-Actuated Control of a Mobile Robot by Human EEG. ieee transactions on biomedical engineering, vol. 51, no. 6, june 2004 1026-1033

[11] Hyun K. Kim , S. James Biggs, David W. Schloerb, Jose M. Carmena,,Mikhail A. Lebedev, Miguel A. L. Nicolelis, and Mandayam A. Srinivasan. Continuous Shared Control for Stabilizing Reaching and Grasping With Brain-Machine Interfaces. IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 6, JUNE 2006.1164-1173.

[12] F. Gala n, M. Nuttin, E. Lew, P.W. Ferrez, G. Vanacker, J. Philips, J. del R. Milla′n. A brain-actuated wheelchair: Asynchronous and non-invasive Brain–computer interfaces for continuous control of robots . Clinical Neurophysiology 119 (2008) 2159–2169

[13] Ying Gu, Kim Dremstrup.Dario Farina. Single-trial discrimination of type and speed of wrist movements from EEG recordings. Clinical Neurophysiology 120 (2009) 1596–1600.

[14] Ying Gu, Dario Farina1, Ander Ramos Murguialday , Kim Dremstrup1, Pedro Montoya and Niels Birbaumer. Off line identifi cation of imagined speed of wrist movements in paralyzed ALS patients from single-trial EEG .Frontiers in Neuroprosthetics. August 2009 .Volume 1 ,Article 3,1-7.

[15] Han Yuan, Christopher Perdoni and Bin He Relationship between speed and EEG activity during imagined and executed hand movements J. Neural Eng. 7 (2010) 026001 1-10.

[16] Hochberg L R, Serruya M D, Friehs G M, et al. Neuronal ensemble control of prosthetic devices by a human with tetraplegia [J]. Nature, 2006,442 (13 ) :164-171.

[17] Chapin J K, Moxon K A, Markowitz R S ,eta1.Real-time control of arobot arm using simultaneously recorded neurons in the motor cortex [J].Nature , 1999,2( 7 ) :664-670

[18] G. Vanacker, J.d.R. Millán, E. Lew, P.W. Ferrez, F. Galán, J. Philips, H. Van Brussel, and M. Nuttin, “Context-based filtering forassisted brain-actuated wheelchair driving,’’ Comp Intell Neurosci, 2007

[19] Tallon-Baudry, C.& Bertrand, O. (1999),Oscillatory gamma activity in humans and its role in object representation. Trends in Cognitive Science,3,151-162..

[20] WeiJingHan,YanKeLe, etc. The cognitive neuroscience. May 2008 ,the irst edition, people's education press. 124-127，140-141

[21] Yunfa Fu etc..Event-related Perturbation in Spectral Power and in Potentials during Periodic Fast and Slow Motor Imagination for Brain-controlled Robots Interface. Proceedings of the 2010 IEEE International Conference on Robotics and Biomimetics December 14-18, 2010, Tianjin, China,1299-1304

[22] Yunfa Fu etc.Time Domain Features for Relationship between Speed and Slow Potentials Activity during Periodic Movement and Motor Imagery at Fast and Slow for BCRI. 2011 3rd International Conference on Bioinformatics and Biomedical Technology,Submitted.

[23] Arnaud Delorme, Scott Makeig EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods 134 (2004), 9-2.

[24] http://thebrain.mcgill.ca/flash/i/i_06/i_06_cr/i_06_cr_mou/i_06_cr_mou.html

[25] http://brainconnection.positscience.com/topics/?main=gal/motor-cortex [26] GERT PFURTSCHELLER AND CHRISTA NEUPER. Motor Imagery

and Direct Brain-Computer Communication. Invited Paper,IEEE,VOL.89.NO.7,JULY 2001,1123-1134.

[27] Pfurtscheller G and Lopes da Silva F H 1999 Event-related EEG/MEG synchronization and desynchronization: basic principles Clin. Neurophysiol. 110 1842–57.

[28] Pfurtscheller, Gert; Andrew, Colin Event-Related Changes of Band Power and Coherence: Methodology and Interpretation Journal of Clinical Neurophysiology: November 1999 - Volume 16 - Issue 6 - p 512

[29] Neuper C, Wörtz M, Pfurtscheller G. ERD/ERS patterns reflecting sensorimotor activation and deactivation. Prog Brain Res. 2006;159:211-22