Essay details:

  • Subject area(s): Engineering
  • Price: Free download
  • Published on: 7th September 2019
  • File format: Text
  • Number of pages: 2

Text preview of this essay:

This page is a preview - download the full version of this essay above.

Electroencephalographic Analysis of Cortical Activity in Selective Attention and Short-Term Working Memory Task using Multiscale Principal Component Analysis

Authors: Jupitara Hazarika, Piyush Kant, Rajdeep Dasgupta, Sahedul Haque Laskar

ABSTRACT: To selectively attend received information (selective attention) inhibiting the unwanted part and to retain the information for a short duration of time (short term working memory) are important cognitive phenomena. Though both are interrelated, there is a lack of evidence to prove the hypothesis. In the present study, we have investigated the electroencephalographic activity in alpha and gamma frequency range reflecting the visual information processing in selective attentional control and non-verbal short term working memory task. Mean and energy values have been computed from both the frequency components and are employed as features. An algorithm is proposed using Multiscale Principal Component Analysis (MSPCA) denoising method and K-Nearest Neighbor (KNN) classifier for the analysis purpose. Our experimental findings reveal that the classifier gives 78% to 96.9% accuracy when different feature vectors are compared.

Keywords: EEG, Multiscale PCA, Selective attention, short term working memory


Visual attention is the major area of investigation in the field of cognitive neuroscience and related research. It involves careful concentration on a single object or thought in a multi-stimulus environment of the vision system. For the healthy cognitive functioning of the brain including learning, decision making, reasoning, problem-solving, perception etc., the visual information processing is an essential phenomenon. The selective attention is the ability of the brain to select the relevant information from many input stimuli while filtering out the other distractions {Chang, 2011; Diamond, 2013}. The limited capacity of the vision system limits the processing of whole information received via retina. Hence the brain relies on the attention to bring forward the important details and to suppress the background information. Inhibitory control makes it possible.

Visuospatial short-term working memory (STWM), on the other hand, relies on attentive selection. It defines the ability to store received visual information in mind for a short period of time without manipulation {Diamond, 2013}. It keeps the information in the readily available state for temporary recall {Cowan, 2009}. The requirement of focused attention on the attained information might infer that selective visual attention and visuospatial short-term WM are related {Diamond 2013}. Although it is controversial that the neural mechanism underlying these two functions are similar or not.

To resolve the controversy, we analyzed the Electroencephalogram (EEG) recorded while subjects were performing a selective attention task and a visuospatial STWM task.  The use of EEG measures in analyzing brain dynamics to understand the cognitive behavior is one of the conventional methods. This non-invasive technique of recording the electrical activity over the scalp of the brain with very high temporal resolution has made it possible to understand the underlying neural activity responsible for a different kind of information processing. The typical classification of EEG spectrum is based on the frequency content of the signal which includes: delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12-30 Hz) and gamma (above 30Hz). Literature suggests the highest involvement of the alpha and gamma frequency ranges in visual attentional control and in STWM function.

The alpha frequency range is found to be dominant in the human scalp of adults during awake, resting but conscious condition {Eidelman-Rothman, 2016; Kay, 2012; Klimesch, 1999}. Initially, the alpha frequency range has been considered to be associated with ‘cortical idling’ {Pfurtscheller, 1996} i.e. task related inactivity. However, the recent evidence has demonstrated its inhibitory function in the task-irrelevant cortical areas {Jensen, 2010; Klimesch, 2012; Hummel, 2002}. Decreased alpha activity (desynchronization) has been reported in the engaged areas and increased alpha activity (synchronization) has been reported in disengaged areas of the brain {Doesburg, 2009; Rihs, 2007}. This specifies the significance of alpha synchronization in both selective visual attention task {Rihs, 2007; Klimesh, 2012} as well as STWM retention tasks {Haegens, 2010}.

Conversely, the effect of information processing is reverse on the high-frequency EEG component. Neural activity in gamma oscillations (>30Hz) is proposed to be a fundamental process in cortical computation (Fries, 2009) like sensory processing {Jensen et al., 2007; William Sedley and Mark O. Cunningham, 2013}. The behavioral gamma response has been studied in different selective visual attention demanding tasks {Ray et al., 2008; Juan et al. 2006; Sateri, 2013; Kahlbrock, 2013; Alex Goddard, 2012} and STWM tasks {Lewis, 2005; Howard, 2003}. The synchronized activity in the gamma band has been found to reflect the active involvement of the engaged network {Fries, 2016; Jensen, 2010}.

Though the scalp EEG provides high-resolution information of the ongoing brain’s activity, the recording of EEG includes noise contaminations. The major artifacts include eye movement, eye blinks, electrocardiogram (ECG), muscle movement and powerline interference {Jung, 2000}. Eye movement and eye blinks are common artifacts in EEG data acquisition. It generates a potential difference between the cornea and the retina known as Electrooculogram (EOG) in the frequency range of 1–3 Hz (typically in the delta frequency range) and it most strongly affects the frontal lobe {Mammone, 2012}. Also, the muscle artifact contributes to higher frequency bands {Zhou, 2005}. To denoise the raw EEG signals, the use of wavelet-based Multiscale Principal Component Analysis (MSPCA) has been shown recent development {Sharma, 2010; Gokgoz, 2014: Alickovic, 2015;  Mahajan , 2015}. It is a nonlinear filtering approach reducing the decorrelation of the correlated data set in each stage of the wavelet transform.

Hence, we denoised the recorded EEG signals using MSPCA and examined the simultaneous engagement of the alpha and gamma frequency features, in the execution of the selective attentional task and STWM task. We, also, propose an algorithm employing MSPCA and KNN to classify the neurophysiological involvement of healthy human in mentioned tasks.



We analyzed data from nine (9) right-handed male young healthy adults, ages between 25 to 29 years (mean age= 27.8 years). Participants had declared that they are free from any neurological disorders, sleep disorder, psychiatric diseases, or any medical conditions.

2.2. Procedure and Stimuli:

For the test, Psychology Experiment Building Language (PEBL) test battery {Mueller & Piper, 2014} software is installed on a computer with 18.5-inch LED monitor and participants are asked to perform the tests. A 2-minute baseline period was incorporated before starting the test and in between the two tests- Bivalent Shape Task (BST) and Corsi Block Tapping Test (CBT). The experiment was performed in a quiet room during daylight.

Bivalent Shape Task (BST): It is a non-verbal task where a test shape (a circle or a square) appears at the center of the screen and the participant needs to confirm the matching shape, ignoring the color and the size [Figure 1]. It is a Stroop task involving the selective attention with inhibitory control {Mueller, S. T., & Esposito A. G., 2014}.

Corsi Block Tapping Test (CBT): In each trial, the task involves nine (9) blocks on the screen and randomly, they lit up one by one in the computer version [Figure 2]. The participant needs to remember the sequence and mimic the sequence in the same order as they were lighted. The task initiates with lighting three blocks per trial and it increases once the participant replies with two correct sequences.

2.3. EEG data acquisition:

While sitting on a comfortable chair in front of the monitor, the EEG data were recorded from each individual participant with an electrode cap from the locations: Fp1, Fp2, F3, F4, F7, F8, Fz, C3, C4, Cz, P3, P4, P7, P8, Pz, O1, O2, Oz, referenced to mastoid. The international 10–20 EEG electrode placement system {Jasper, 1958} was followed to identify proper electrode positions [Figure 3] and BIOPAC Mobita 32-channel wireless EEG system was used for acquiring EEG data. Electrode impedances were maintained below 5kΩ and the sampling frequency was 2 kHz.

An initial 2-minutes eye open baseline recording was performed during which the participants were asked to concentrate on a blank screen. During PEBL tests, written instructions appeared on the screen at the beginning of each test and the participants were trained on the task by making them try it for once before data acquisition. A 2-minute resting period was there in between the mentioned tasks.

EEG data analysis:

The recorded EEG data were analyzed in MATLAB (R2015a) software. During each recording period, the subjects were asked to minimize the eye blink and muscle movements.

The high sampling rate at the time of acquisition yields large data files which might slow down the computation process. Therefore, the sampling rate was reduced to 256Hz by resampling the data and according to the Nyquist rule, the highest frequency present in the signal became 128Hz. A 50Hz Butterworth Notch filter was used to remove the powerline interference. To eliminate the other noise contaminations, the wavelet-based Multiscale PCA (MSPCA) was used.

2.4.1. Multiscale Principal Component Analysis (MSPCA):

Multiscale Principal Component Analysis (MSPCA) combines the facilities of PCA with the wavelet transform. The PCA is a linear orthogonal transformation that finds a linearly uncorrelated set of variables, called principal components, from the set of correlated variables. The MSPCA performs PCA on the approximate as well as on each detailed coefficients of Discrete Wavelet transform (DWT) {Bakshi, 1998}.

In DWT, discretely sampled wavelets are used and it can be defined as follows {Handojoseno et al., 2015}:

DWT(j,k)=1/√(|2^j | ) ∑_(t=-∞)^∞▒〖x(t)ψ((t-2^j k)/2^j )〗…………………….(2)

Where the parameters a and b are replaced by 2^jand 2^j k respectively.

DWT decomposes a discrete signal x[n] into detail and approximation wavelet coefficients by passing it through a high pass filter h[n] and a low pass filter g[n] simultaneously, where both the filters have the same cut-off frequency which is half of the frequency of x[n].

Hence the low pass and high pass filter outputs are given by

x_low [n]=x[n]*g[n]=∑_(k=-∞)^∞▒〖x[k]g[n-k]〗……………………….(3)

x_high [n]=x[n]*h[n]=∑_(k=-∞)^∞▒〖x[k]h[n-k]〗……………………….(4)

Where g[n] and h[n] are the impulse responses of the respective filters.

The filtered outputs are then down-sampled by 2 as according to Nyquist’s rule, half the samples can be discarded since half the frequencies of the signal x[n] have been removed. Thus the approximate and detail coefficients, after one level of decomposition, can be mathematically expressed as:

x_(Approx.) [n]=(x[n]*g[n] )↓2=∑_k▒x[k]g[2n-k]                          (5)

x_(Detail.) [n]=(x[n]*h[n] )↓2=∑_k▒x[k]h[2n-k]                              (6)

Where, x_(Approx.) [n] and x_(Detail.) [n] are the approximate and detail output coefficients of DWT 1-level decomposition [Figure 4].

Let us consider n×m sized matrix X, where m is the number of signals (waveforms) each of length n. Each signal (column) in X is decomposed using the wavelet transform into the wavelet coefficients. The matrix W holds the wavelet coefficients gm and hj [Bakshi, 1998]:

W=〖[H_j      G_j       G_(j-1)……..G_p……….G_1]〗^T                                 (7)

The following steps of the MSPCA are illustrated in Figure 5 and explained by the following algorithm [Jasmin Kevric & Abdulhamit Subasi 2014]:

1) Calculate wavelet decomposition WX at scale level j for each column in data matrix X,

2) For 1≤p≤j, execute the PCA of the detail matrices G_p X and choose a suitable number of significant principal components or reject the detail,

3) Execute the PCA of the approximation matrix H_j X and choose a suitable number of principal components,

4) By inverting the wavelet transform〖 X〗^T, recover a new matrix from the reduced detail and approximation matrices,

5) Lastly, execute the PCA of that new matrix to form.

 In the present study, 5-level 4th order Daubechies (db4) was used to each EEG signal and a suitable number of significant principal components were retained using the Kaiser’s rule. The Kaiser’s rule keeps the components associated with eigenvalues exceeding the mean of all eigenvalues.

2.4.2. Wavelet Packet Decomposition:

Wavelet packet decomposition is another form of wavelet transform where both the low pass filter output i.e. the approximate coefficients and the high pass filter output i.e. the detail coefficients of the original signal are further subdivided into finer frequency bands by using two-scale relations repeatedly. Hence provides better time-frequency resolution. A wavelet packet tree is as shown below:

In the present study, each EEG signal, after MSPCA denoising, was decomposed using wavelet packet decomposition with 5-level 4th order Daubechies (db4) to extract the alpha (8-12 Hz), gamma1 (32-64Hz) and gamma2 (64-128Hz) frequency components from the wavelet coefficients [5, 2], [2, 2] and [1,1] respectively.

3. Spectral Feature extraction:

Mean and energy was calculated out as features from each frequency component to assess the neuropsychological involvement in the tasks.

Mean: In statistics, the arithmetic mean of a dataset refers to the center value of distribution. For a set of numbers x_1,x_2,……………….x_n, the arithmetic mean is calculated as:

x ̅=1/n ∑_(i=1)^n▒〖x_n                                                            (8)〗

Energy: In case of signal processing, the energy of a discrete signal x[n] is given by

E=∑_(n=1)^n▒|x[n]|^2                                                        (9)

Thus, the percentage relative band energy of ith frequency band is given by

E_i=E_i/E_total ×100%                                               (10)

Where, E_total= Energy of the original signal.

2.4.4. K-nearest neighbors (KNN) algorithm:

The k-nearest neighbors (KNN) is a simple classification algorithm based on the assumption that class probabilities are locally approximately constant. For a user defined value of K, the algorithm finds the K-nearest neighbor samples from the available training data set closest in distance to the new point and predicts the class from majority voting. Standard Euclidean distance is the most common choice in this case to calculate the distance.

For test instance x_i  , the Euclidean distance from a neighbor instance sample y_i is calculated as:

D=√(∑_(i=1)^K▒(x_i-y_i )^2 )                                                (11)

In this experiment, we tried 3, 4 and 5 for K-value and finally selected 5 for the classifier as it gave maximum accuracy.


Observation of the brain’s dynamic activity in selective attentional control and in working memory task demonstrated the participation of prefrontal, frontal, parietal and occipital lobe in both the tasks.

On comparing with the baseline, the mean value and the band energy of alpha showed an increment during the Bivalent shape task (BST) at positions F7, F8, and at all the center positions - Fz, Cz, Pz, POz and Oz (Figure 7 & Figure 8). However, both the features of alpha frequency band decreased at almost all the electrode positions, except the energy at F8, in the case of Corsi block test (CBT) when compared with the baseline condition (Figure 7 & Figure 8). On comparing BST and CBT tasks, the alpha activity of BST was lower than that of CBT in the locations of Fp2, C3, P4 in the case of mean value, but only at C3 location in case of band energy.

The gamma1 and gamma2 increased at all the locations in both the tasks compared to the baseline, considering mean values averaged over all the participants (Figures 9, 10, 11, 12). Mean values of gamma1 were either negative or near to zero at all the positions whereas mean values of gamma2 were negative for all the electrode positions during the baseline. Desynchronization in gamma1 band energy was observed at occipital position O1 during BST, compared to the baseline and in gamma2 band energy at C3 during CBT, compared to BST. In all other electrode positions, both the features of both high gamma and low gamma oscillations were synchronized when BST and CBT activities were compared to the baseline.

The extracted features from the three frequency bands were used to construct the training and testing sets of 5-fold KNN classifier. For each frequency band, two features from each channel constituted a dataset of 9subjects×2 features×21channels for each task. Out of which 60% data were used to train the classifier and 40% data were used for testing purpose. The classification accuracies of the algorithm for different frequency bands with extracted features are displayed in table1.


The findings of this study demonstrated that selective attention and short-term working memory has a different impact on low and high-frequency response. The synchronization in alpha is an index of active inhibitory function in task-irrelevant areas and desynchronization represents active participation in the task. Hence the increased midline alpha showed an active inhibitory function along with two frontal locations F7and F8, during the selective attentional task {Ghimire, 2014}. But the short-term WM task showed inhibitory function only at the F8 location.

The increased gamma activity has been related to active information processing in engaged locations. Higher power in the gamma band in frontal, parietal, temporal and occipital lobe were reported in previous studies while performing the selective attentional task {Gruber et al., 1999} and in the STWM task {Roux, 2014; Jokisch, 2007; Tallon-Baudry, 1998}. This is consistent in our study as increased activity in both gamma1 and gamma2 frequency bands were reported in the BST task indicating increased information processing in selectively attempting the target.  In the STWM task, increased gamma activities might be due to successful recall {Jokisch,, 2007} and memory load {Howard, 2003}.

Exceptionally, at C3 position, the alpha band (8-12Hz) showed increment and the higher gamma band (64-128Hz) showed decrement during STWM task than the selective attention task which might suggest that active inhibition was released at this area in the STWM task.

The negative mean values found in gamma1 and gamma2 frequency bands might refer the cortical inactivity during baseline condition. The lower energy values in high gamma as well as in low gamma, in the motor cortex areas, might signify that the involvement of this area is low in visual information processing.

Furthermore, it is observed from table-1 that the classification accuracies are quite high in all cases. On comparing BST with CBT, alpha activities showed 87.5% and both gamma1 and gamma2 activities showed 96.9% classification accuracies. When both the features from all the three frequency bands were used in the KNN classifier, it gave 89.6% accuracy. This suggests that the cortical activities in selective attention task and STWM task are significantly different from each other.


Previous research has studied visual selective attention and short-term working memory (STWM) as overlapping constructs. This paper aimed to visualize the cortical activities referring these two cognitive tasks in terms of alpha and gamma frequency bands and to propose a promising classification algorithm. The findings of this paper support the task-relevant involvement of frontal, post-parietal and occipital locations in alpha and gamma frequency bands. The acquired EEG signals were first denoised using wavelet-based MSPCA technique, decomposed into frequency bands using wavelet transform and then the features extracted from the signals were fed to KNN classifier. Classification accuracies were found to be quite high in all cases. Hence it can be stated that dimensionality reduction using PCA at the wavelet decomposed levels reduces the artifact and helps in keeping the task-relevant information in the signal. Further investigation on EEG-based cognitive performance analysis using different feature extraction and classification techniques may give further insight.


J. U. Duncombe, “Infrared navigation—Part I: An assessment of feasibility (Periodical style),” IEEE Trans. Electron Devices, vol. ED-11, pp. 34–39, Jan. 1959.

M. Y. Chang, R. S. Dean, “Selective Attention,” Encyclopedia of Child Behavior and Development, pp 1300-1301, 2011.

Damond 2013

N. Cowan, “What are the differences between long-term, short-term, and working memory?,” Progress in Brain Research, vol.169, pp. 323–338, 2008.

W. Klimesch, “Alpha-band oscillations, attention, and controlled access to stored information,” Trends in Cognitive Science, vol. 16(12), pp. 606–617, Dec. 2012. doi:  10.1016/j.tics.2012.10.007

W. Klimesch, “EEG alpha and theta oscillations reflect cognitive and memory performance: A review and analysis,” Brain Research Reviews, vol. 29, pp. 169-195, 1999.

W. Klimesch (2002), “Alpha-band oscillations, attention, and controlled access to stored information,” Trends in Cognitive neuroscience, Vol. 16, Issue 12, December 2012, Pages 606–617,

O. Jensen, J. Kaiser and J. P. Lachaux, “Human gamma-frequency oscillations associated with attention and memory,” Trends in Neuroscience, vol. 30(7), pp. 317–324, July 2007.

O. Jensen and A. Mazaheri, “Shaping functional architecture by oscillatory alpha activity: gating by inhibition,” Frontiers in Human Neuroscience, 4, 2010. 10.3389/fnhum.2010.00186

O. Jensen, J. Gelfand, K. Kounious and J. E. Lisman, “Oscillations in the alpha band (9–12 Hz) increase with memory load during retention in a short-term memory task,” Cereb Cortex, vol. 12 (8), pp. 877–882, August 2002.

F. Hummel, F. Andres, E. Altenmüller, J. Dichgans and C. Gerloff, “Inhibitory control of acquired motor programmes in the human brain,” Brain, vol. 125, pp. 404–420, Feb. 2002.

S. M. Doesburg, J. J. Green, J. J. McDonald and L. M. Ward, “From local inhibition to long-range integration: A functional dissociation of alpha-band synchronization across cortical scales in visuospatial attention,” Brain Research, vol. 1303, pp. 97–110, Dec. 2009.

T. A. Rihs, C. M. Michel and G. Thut, “Mechanisms of selective inhibition in visual spatial attention are indexed by α-band EEG synchronization,” European Journal of Neuroscience, vol. 25 (2), pp. 603–610, Jan. 2007.

Haegens, S. et al. (2010) Somatosensory working memory performance in humans depends on both engagement and disengagement of regions in a distributed network. Hum. Brain Mapp. 31, 26–35

Friese, U. et al. Oscillatory brain activity during multisensory attention reflects activation, disinhibition, and cognitive control. Sci. Rep. 6, 32775; doi: 10.1038/srep32775 (2016).

Fries P. Neuronal gamma-band synchronization as a fundamental process in cortical computation, Annu Rev Neurosci, 2009, vol. 32, (pg. 209-24).

William Sedley and Mark O. Cunningham, Do cortical gamma oscillations promote or suppress perception? An under-asked question with an over-assumed answer, Frontier of Human Neuroscience, 20 September 2013,

Supratim Ray, Ernst Niebur, Steven S. Hsiao, Alon Sinai, Nathan E. Crone, (2008). High-frequency gamma activity (80–150 Hz) is increased in human cortex during selective attention, Clinical Neurophysiology, Volume 119, Issue 1, January 2008, Pages 116–133.

Juan R. Vidal, Maximilien Chaumon, J. Kevin O\'Regan, and Catherine Tallon-Baudry (2006), Visual Grouping and the Focusing of Attention Induce Gamma-band Oscillations at Different Frequencies in Human Magnetoencephalogram Signals, Journal of Cognitive neuroscience, November 2006, Vol. 18, No. 11, Pages: 1850-1862

Rouhinen, Sateri; Panula, Jonatan; Palva, J. Matias; Et Al., Load Dependence Of Beta And Gamma Oscillations Predicts Individual Capacity Of Visual Attention, Journal Of Neuroscience   Volume: 33   Issue: 48   Pages: 19023-19033   Published: Nov 27 2013

Kahlbrock, N.; Butz, M.; May, E. S.; et al., Gamma Band Oscillations are related to the Degree of selective visual attention, KLINISCHE NEUROPHYSIOLOGIE   Volume: 44   Issue: 1   Pages: 18-23   Published: MAR 2013

C. Alex Goddard, Devarajan Sridharan, John R Huguenard, Eric I Knudsen (2012), Gamma Oscillations Are Generated Locally in an Attention-Related Midbrain Network, Neuron 73(3):567-80 • February 2012

Cortical inhibitory neurons and schizophrenia, DA Lewis, T Hashimoto, DW Volk - Nature Reviews Neuroscience, 2005 -

Marc W. Howard, Daniel S. Rizzuto2, Jeremy B. Caplan, Joseph R. Madsen2,3, John Lisman2, Richard AschenbrennerScheibe4, Andreas Schulze-Bonhage4 and Michael J. Kahana (2003). Gamma Oscillations Correlate with Working Memory Load in Humans, Cerebral Cortex December 2003;13:1369–1374; DOI: 10.1093/cercor/bhg084

Jung, T.-P., Makeig, S., Humphries, C., Lee, T.-W., McKeown, M. J., Iragui, V. and Sejnowski, T. J. (2000), Removing electroencephalographic artifacts by blind source separation. Psychophysiology, 37: 163–178. doi:10.1111/1469-8986.3720163

L. N. Sharma; S. Dandapat; A. Mahanta (2010). Multiscale principal component analysis to denoise multichannel ECG signals,2010 5th Cairo International Biomedical Engineering Conference, Pages: 17 - 20, DOI: 10.1109/CIBEC.2010.5716093

Gokgoz, E. & Subasi, A. (2014), Effect of multiscale PCA de-noising on EMG signal classification for diagnosis of neuromuscular disorders, J Med Syst, 38: 31. doi:10.1007/s10916-014-0031-3

Alickovic, E. & Subasi, A. (2015) Effect of Multiscale PCA De-noising in ECG Beat Classification for Diagnosis of Cardiovascular Diseases, Circuits Syst Signal Process 34: 513. doi:10.1007/s00034-014-9864-8

Ruhi Mahajan ;  Bashir I. Morshed (2015) Unsupervised Eye Blink Artifact Denoising of EEG Data with Modified Multiscale Sample Entropy, Kurtosis, and Wavelet-ICA, IEEE Journal of Biomedical and Health Informatics ( Volume: 19, Issue: 1, Jan. 2015 ), Page(s): 158 – 165, DOI: 10.1109/JBHI.2014.2333010

B. R. Bakshi, “Multiscale PCA with Application to Multivariate Statistical Process Monitoring,” AlChE, vol. 44, no. 7, pp. 1596-1610, 1998.

Mueller, S. T., & Esposito A. G. (2014). Computerized Testing Software for Assessing Interference Suppression in Children and Adults: The Bivalent Shape Task (BST). Journal of Open Research Sofware, 2(1).

Handojoseno, A. M. A., Shine, J. M., Nguyen, T. N., Tran, Y., Lewis, S. J. G., & Nguyen, H. T. (2015). Analysis and Prediction of the Freezing of Gait Using EEG Brain Dynamics. Ieee Transactions on Neural Systems and Rehabilitation Engineering, 23(5), 887-896.

Nisha Ghimire, Bishnu Hari Paudel2, Rita Khadka3, Paras Nath Singh, Asim Das, Electroencephalographic changes during selective attention (2014), Asian Journal of Medical Sciences | Mar-Apr 2015 | Vol 6 | Issue 2,pp-51-56.

Thomas Gruber, Matthias M. Müller, Andreas Keil, Thomas Elbert (1999), Selective visual-spatial attention alters induced gamma band responses in the human EEG, Clinical Neurophysiology, Volume 110, Issue 12, 1 December 1999, Pages 2074–2085

F Roux, PJ Uhlhaas  (2014), Working memory and neural oscillations: alpha–gamma versus theta–gamma codes for distinct WM information? - Trends in cognitive sciences, 2014 - Elsevier

Jokisch, D. and Jensen, O. (2007) Modulation of gamma and alpha activity during a working memory task engaging the dorsal or ventral stream. J. Neurosci. 27, 3244–3251

Tallon-Baudry, C. et al. (1998) Induced gamma-band activity during the delay of a visual short-term memory task in humans. J. Neurosci. 18, 4244–4254.

...(download the rest of the essay above)

About this essay:

This essay was submitted to us by a student in order to help you with your studies.

If you use part of this page in your own work, you need to provide a citation, as follows:

Essay Sauce, . Available from:< > [Accessed 29.05.20].