reference. Unfortunately, on the scalp positions, there is no electrode with zero potential.
Therefore, we usually take the right ear as the reference known as monopolar recording.
This means that the measured EEG signals are depended. To convert reference-depended signals to reference-free signals, we can apply reference-independent derivation to the reference-dependent signals such as Common average reference (CAR), Bipolar ref-erence, Laplacian reference. We will give more details about the three approaches in Chap-ter 4.
1.2.3 Brain signal
Neurons can send action potentials, postsynaptic potentials to other neurons by through the dendrites and axons. An action potential is initiated when the voltage is larger than the threshold about -40 mV. In fact, what EEG measures are the potentials integrating from the thousands or billions synchronously activated neurons but not the potential of one neuron for it is too small to be detected. The recorded EEGs include spontaneous electrical activity of the cerebral cortex and the cortical response evoked by external or internal events, namely event-related potential. In general, it is believed that what EEGs
1.2 Background 7
Table 1.2: Common brain rhythms in different frequency bands.
Band. Frequency [Hz]
Delta < 3.5 Theta 4-7.5
Alpha 8-13
Beta >13
measures is the summation of excitatory and inhibitory postsynaptic potentials of cerebral cortex with some contribution of granular and glia cell activity [27].
There exists three typical processing techniques of the measured EEGs, which are com-monly used in BCI systems:
• Brain rhythm
• Event-related potential(ERP)
• Event-related desynchronization(ERD) and event-related synchronization(ERS) Brain rhythm and ERD/ERS are based on analysis of spontaneous brain activity while ERP on cortical responses to events.
Brain Rhythm
Human brain has various different rhythmic activities according to the different fre-quency range. Typical rhythmic activities include delta, theta, alpha, beta and gamma rhythms. We indicate the relationship between the rhythms and frequency ranges in ta-ble 1.2. In Figure 1.5, an example for these brain rhythms [25] is illustrated.
In this thesis, we only concern the alpha (mu) and beta rhythms because they will react to the imagination of hand movement which is used as the predefined mental tasks in this work. The alpha, mu and beta rhythms are described as follows:
1. Alpha rhythm Alpha rhythm is between 8-13 Hz with amplitude mostly below 50uV for adults. Alpha rhythm is considered as the primary rhythm of normal adult brain.
Alp ha Beta
T hed a
Delta
Figure 1.5: An example of four brain rhythms [21].
It is frequently seen in occipital areas when people is awake , close their eyes and under conditions of physical relaxation [45].
2. Mu rhythm Mu rhythm is in the aplha frequency band. The main difference between mu and alpha rhythm is that the mu rhythm is usually blocked or attenuate with con-tralateral movements or the thought of a movement, but not react to eye open or closing. For example, when executing a movement, the central mu rhythm is desyn-chronized; however, the occipital alpha rhythm is synchronized. Figure 1.6 shows a contralateral localized mu event-related desynchronization (ERD) and a occipital lo-calized alpha rhythm event-related synchronization (ERS) in a right hand movement task. In the next section we will give an overview of event-related desynchronization/event-related synchronization ERD/ERS.
During a movement, the elicited contralateral mu rhythm is interpreted as an unspe-cific presetting, priming of neurons in motor areas. Besides motor-relative tasks, the mu rhythm can also be induced by a flicker stimulation [41], a reading task [37].
3. Beta rhythm The beta rhythm is defined as a frequency of above 14Hz and below 30Hz, with amplitude usually being seldom larger than 30uV and irregular. Beta rhythm is also the background feature of normal adult brain. It can be found over the frontal and cerebral regions. The central beta rhythm can be elicited when people perform a
1.2 Background 9
ERD
ERS
Figure 1.6: The contralateral localized mu ERD and a occipital localized alpha rhythm ERS in a right hand movement task is illustrated. The topography is at about the movement execution moment [36].
voluntary movements.
Event-related potentials(ERPs)
Several kinds of internally or externally paced events will result in time-locked and phase-locked brain signals. Almost all of these kinds evoked activities have a more or less fixed time-delay to the stimulus. These time-locked and phase-locked brain signals are called event-related potentials (ERP) or evoked potentials (EP). ERP can be viewed as potential changes of the neurons when our brain deal with mental tasks.
Usually the brain signals of a mental task is smaller than the ongoing brain signals, thus concealed in the irregular and noisy ongoing brain signals. In order to extract the ERPs, synchronous averagingare performed, implying we have to execute the same mental tasks more than once. Due to the non-time-locked and non-phase-locked of the noise, after applying synchronous averaging, most of the noise will be eliminated, therefore enhancing the signal-to-noise ratio and obtaining the time-locked and phase-locked signals, ERP.
Many various ERPs have been proposed today such as slow cortical potential (SCP), P300, visual evoked potential (VEP), and steady-state visual evoked potential (SSVEP).
The way to label ERPS is often the latency and the electrical polarity or the type of given stimuli. An example of the former is P300, representing a positive peak with latency 300ms
t op t ar get
bot t om t ar get
-7m V
+7m V
Ti m e(s)
amplitude (mV)
Figure 1.7: An example of SCP when the user performs a task that move a cursor toward a target at the bottom or the top [57].
posterior to the stimulus onset. For the latter, the representative ERP is visual event-related potential(VEP), elicited by a visual stimulus. Here, the details of these four ERPs, related to some existing BCI systems, are discussed.
• SCP:
SCPs are the slow voltage changes of the brain cortex, with a 0.5-10.0s potential shifts. They are settled in the frequency range below 1-2Hz. SCPs can be divided into two types, negative and positive. Negative SCPs represent the mobilization or readi-ness while positive SCPs represent ongoing congnitive and inhibition of neuronal activity [18]. In 1982, Lutzenberger gave an example that the subject could solve arithmetic problems faster after producing a negtive SCP [18]. Figure 1.7 shows an example of the SCP when the user performs a task moving the cursor toward a target at the bottom or the top. As we see, bottom target often induce positive SCPs while top target negative SCPs. In this figure, we can also find that the SCP persist many seconds.
• P300:
Infrequent, particular or oddball stimuli in auditory, visual or somatosensory will evoke the P300. P300 is a positive peak reaching the maximum of about 300ms after
1.2 Background 11
o ther c ho ic e d es ired
c ho ic e
P z
Figure 1.8: An example of P300 in a oddball visual stimuli experiment. [57]
the stimulus over the parietal areas. P300 is commonly utilized in a spelling BCI system [15, 35]. Because the P300 is a naive response to an infrequent stimuli, the user requires no user training to produce a P300 pattern.
• VEP:
Visual evoked potential (VEP) is induced when the user’s eyes are stimulated by looking at a test pattern which often is a flashing pattern. To measure VEPs, the recording electrodes are placed over the visual cortex.
• SSVEP:
The SSVEP is a brain potential changes elicited by a brief visual stimulus modulated at a specific frequency. The SSVEP is charachterized as an increase in EEG activity at the stimulus frequency.
Event-related desynchronization(ERD) and event-related synchronization(ERS) Since Berger(1930), we have known some of our brain signals could block or desyn-chronize the ongoing brain activity. Because these types of signals are time-locked but not
phase-locked, we cannot use the simple linear method such as averaging to extract them, but use frequency analysis.
In 1977, Pfurscheller proposed the concepts of event-related desynchronization(ERD) and event-related synchronization(ERS). He thought these types of brain signals represent specific changes of the ongoing brain activity and might consist, in general, of decrease or of increase of the power in a specific frequency bands. He defined the former as ERD and the latter as ERS. Different from the ERP, which can be thought as a series of changes of the post-synaptic response, the ERD/ERS can be considered as the controlling of brain oscillations.
The ERD/ERS is a percentage value to the power of a predefined interval signals, ref-erence or baseline period. Usually the reference period is an interval before the target is performed. The typical algorithm of calculating ERD/ERS, ERD inter-trial variance), is as follows [36]:
1 Specify the frequency bands we interest and afterward apply bandpass filtering.
2 Calculate the mean of the filtered data over all trials.
3 Subtract the mean from the filtered data in step1.
4 Square the amplitude samples from the previous step over all trials.
5 Average over time samples from the step4.
6 Obtain ERD/ERS by calculating the percentage relative to the power of the reference interval.
These steps can be found in Figure 1.9, which indicates the algorithms of calculating ERP, ERD/ERS. In this figure, there are two approaches we need to take a notice, band-power ERD and ERD inter-trial variance. Both are used to calculate ERD/ERS. The main difference between ERD and ERD inter-trial variance is that the latter subtract the mean of phase-locked signals from the original signals. The motive of performing subtraction is ” a phase-locked power increase due to the ERP can mask the non-phase-locked power decrease(ERD) when the classical band-power method is used.” In Figure 2.2, we present an example of calculating ERD/ERS.
1.2 Background 13
Figure 1.9: The flowchart ERP and ERD/ERS
0 1 2 3 4 5 6 7 8 9 A B Alp ha
Beta 150
100 50 0 -50
power change [%] -100
C 3
Alpha
5.5s
8.5s
Figure 1.10: An example of ERD/ERS. [36]. The mental tasks is imagination of hand movement. The onset time is at 5s. In the left side two ERDs in alpha and beta band are illustrated. In the right side, two topographies are plotted with respect to at 5.5s and at 8.5s.
ERD/ERS can emerge when performing a movement or a thought of movements, per-ceptual, judgement task. For example, Figure 1.11 shows the ERD/ERS in mu bands. The data was measured over the position C3, Cz, C4 for performing hand motor imagery. The contralateral ERD can be found easily in this figure.